Merge model-registry into main

biome.json

@@ -31,6 +31,7 @@
 			"!**/node_modules/**/*",
 			"!**/test-sessions.ts",
 			"!**/models.generated.ts",
+			"!**/*.models.ts",
 			"!packages/mom/data/**/*",
 			"!!**/node_modules"
 		]

packages/agent/CHANGELOG.md

@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+### Breaking Changes
+
+- `AgentHarnessOptions.models` is required and is the only auth path: the harness streams turns, compaction, and branch summarization through the provided `Models` instance (`models.streamSimple()`/`completeSimple()`), resolving auth through the providers. `AgentHarnessOptions.getApiKeyAndHeaders` is removed — apps that resolved keys per request now express that as provider auth (`ApiKeyAuth`/`OAuthAuth`) on the providers in the `Models` collection. Build one with `createModels()` + provider factories (or `builtinModels()` from `@earendil-works/pi-ai/providers/all`); tests use `fauxProvider()`.
+- `compact()`, `generateSummary()`, and `generateBranchSummary()` take a `Models` parameter and no longer accept explicit `apiKey`/`headers`.
+- `StreamFn` is defined structurally (`(model, context, options?) => AssistantMessageEventStream | Promise<...>`); `Models.streamSimple` satisfies it.
+
 ## [0.79.10] - 2026-06-22
 
 ## [0.79.9] - 2026-06-20

packages/agent/docs/models.md

@@ -0,0 +1,918 @@
+# Models architecture
+
+This document describes the target design for the next `pi-ai` model/provider refactor. It describes the desired shape, not the current implementation. It is intended to be complete enough to start implementing from a fresh session.
+
+Goals:
+
+- `Models` is a dumb runtime collection of providers.
+- Concrete providers own metadata, auth, model listing, and stream behavior.
+- API implementations live under `src/api/` and are reusable/lazy.
+- Concrete provider factories live under `src/providers/`.
+- Users can import only the providers they need.
+- Importing a provider must not eagerly import heavy SDKs.
+- Dynamic model lists are first-class: reads are sync (last-known list), fetching happens in an explicit async `refresh`.
+- `models.json` and extensions layer by wrapping providers, not by mutating provider internals ad hoc.
+- Old global APIs survive only in an explicit, temporary `/compat` entrypoint.
+
+Non-goals for the immediate `pi-ai` pass:
+
+- Do not migrate coding-agent `ModelRegistry` yet.
+- Do not keep the stream/API registry inside `Models`.
+- Do not implement web OAuth flows yet.
+- Image generation mirrors the chat-side design (`ImagesModels`/`ImagesProvider` in `images-models.ts`); the old global image API (`images.ts`, `images-api-registry.ts`) lives on compat.
+
+## Package layout
+
+Target source layout:
+
+```txt
+packages/ai/src/
+  index.ts                    # core exports only; no built-in provider imports
+  models.ts                   # Models runtime, Provider
+  images-models.ts            # ImagesModels runtime, ImagesProvider (mirrors models.ts)
+  compat.ts                   # temporary old-API compatibility entrypoint
+  auth/                       # auth method types, helpers, shared resolveProviderAuth(), login callbacks
+  api/                        # API implementations and lazy wrappers
+    openai-completions.ts     # real implementation, imports SDKs, exports stream/streamSimple
+    openai-completions.lazy.ts
+    openai-responses.ts
+    openai-responses.lazy.ts
+    openai-codex-responses.ts
+    openai-codex-responses.lazy.ts
+    azure-openai-responses.ts
+    azure-openai-responses.lazy.ts
+    anthropic-messages.ts
+    anthropic-messages.lazy.ts
+    google-generative-ai.ts
+    google-generative-ai.lazy.ts
+    google-vertex.ts
+    google-vertex.lazy.ts
+    mistral-conversations.ts
+    mistral-conversations.lazy.ts
+    bedrock-converse-stream.ts
+    bedrock-converse-stream.lazy.ts
+    openrouter-images.ts      # image-generation API implementation
+    openrouter-images.lazy.ts
+    lazy.ts                   # lazyStream()/lazyApi() helpers
+    (shared helpers: openai-responses-shared, google-shared, transform-messages, ...)
+  providers/                  # concrete provider factories and per-provider catalogs
+    openai.ts
+    openai.models.ts          # generated OpenAI catalog
+    openai-codex.ts
+    openai-codex.models.ts
+    anthropic.ts
+    anthropic.models.ts
+    google.ts
+    google.models.ts
+    ...one pair per built-in provider...
+    openrouter-images.ts      # image-generation provider factory
+    faux.ts                   # test provider factory
+    all.ts                    # explicit aggregate: builtinModels(), builtinImagesModels(), getBuiltin*()
+  utils/oauth/                # OAuth flow implementations (node), lazy-loaded
+```
+
+`src/index.ts` must stay core-only. It must not import:
+
+- generated model catalogs
+- built-in provider factories
+- provider SDK implementations
+- Node-only OAuth modules
+- `providers/all`
+- `compat`
+
+Provider, API, and compat entrypoints are explicit subpath exports.
+
+## Public usage
+
+Minimal provider usage:
+
+```ts
+import { createModels } from "@earendil-works/pi-ai";
+import { openaiProvider } from "@earendil-works/pi-ai/providers/openai";
+
+const models = createModels();
+models.setProvider(openaiProvider());
+
+const model = await models.getModel("openai", "gpt-4o-mini");
+if (!model) throw new Error("model not found");
+
+const response = await models.complete(model, context);
+```
+
+Multiple providers:
+
+```ts
+const models = createModels();
+models.setProvider(openaiProvider());
+models.setProvider(openrouterProvider());
+```
+
+All built-ins, explicitly heavy metadata entrypoint:
+
+```ts
+import { builtinModels } from "@earendil-works/pi-ai/providers/all";
+
+const models = builtinModels();
+```
+
+`providers/all` may import all provider metadata/catalogs. It still must not eagerly import SDK implementations; provider streams use lazy wrappers.
+
+## Core runtime: Models
+
+`Models` is a provider collection plus auth application and stream convenience. No stream registry, no auth resolver strategy object.
+
+```ts
+export function createModels(options?: {
+  /** App-owned credential storage. Default: in-memory store. */
+  credentials?: CredentialStore;
+  /** Environment access for auth resolution (env vars, file existence). Default: process.env/node:fs backed; injectable for tests and non-Node hosts. */
+  authContext?: AuthContext;
+}): MutableModels;
+
+export interface Models {
+  getProviders(): readonly Provider[];
+  getProvider(id: string): Provider | undefined;
+
+  /** Sync read of last-known models. Best-effort: a provider whose getModels() throws yields no models. */
+  getModels(provider?: string): readonly Model<Api>[];
+  /** Dynamic lists are honestly Model<Api>; narrow with the hasApi() guard. */
+  getModel(provider: string, id: string): Model<Api> | undefined;
+
+  /**
+   * Ask dynamic providers to re-fetch their model lists. With a provider id,
+   * rejects on that provider's failure; without, refreshes all concurrently
+   * best-effort. Static providers are no-ops.
+   */
+  refresh(provider?: string): Promise<void>;
+
+  /**
+   * Resolve request auth for a model. Includes source label for status UI.
+   * Resolves undefined when the provider is unknown or unconfigured. Rejects
+   * with ModelsError ("oauth" on refresh failure, "auth" on api-key/store
+   * failure); status/availability UIs catch rejections and render
+   * "needs re-login" instead of treating them as unconfigured.
+   */
+  getAuth(model: Model<Api>): Promise<AuthResult | undefined>;
+
+  stream<TApi extends Api>(
+    model: Model<TApi>,
+    context: Context,
+    options?: ApiStreamOptions<TApi>,
+  ): AssistantMessageEventStream;
+
+  complete<TApi extends Api>(
+    model: Model<TApi>,
+    context: Context,
+    options?: ApiStreamOptions<TApi>,
+  ): Promise<AssistantMessage>;
+
+  streamSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream;
+  completeSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): Promise<AssistantMessage>;
+}
+
+export interface MutableModels extends Models {
+  /** Upsert/replace by provider.id. Provider ids are unique. */
+  setProvider(provider: Provider): void;
+  deleteProvider(id: string): void;
+  clearProviders(): void;
+}
+```
+
+Removed concepts:
+
+```txt
+no Models.setStreamFunctions() / getStreamFunctions()
+no api-registry as a real dispatch mechanism
+no Models.provider(id) builder, no setModel/upsertModel/patchModel lifecycle
+no ModelAuthResolver / setAuthResolver — resolution policy is fixed, store is injected
+```
+
+If an app needs different auth policy, it wraps providers (wrap auth methods or `getModels`) or passes explicit request auth in stream options.
+
+## Provider
+
+A provider is the concrete runtime unit. It owns id/name/base metadata, auth methods, model listing, and stream behavior.
+
+`Provider` is generic over the APIs its models use. Concrete factories declare what they emit (`openaiProvider(): Provider<"openai-responses" | "openai-completions">`), giving typed model lists to direct factory users. A `Models` collection holds providers as `Provider<Api>`.
+
+```ts
+export interface Provider<TApi extends Api = Api> {
+  readonly id: string;
+  readonly name: string;
+
+  readonly baseUrl?: string;
+  readonly headers?: Record<string, string>;
+
+  /**
+   * Required: at least one of apiKey/oauth. Even ambient-credential providers
+   * (env vars, AWS profiles, ADC) and keyless local servers provide apiKey
+   * auth whose resolve() reports whether the provider is configured.
+   * getAuth() returning undefined = not configured.
+   */
+  readonly auth: ProviderAuth;
+
+  /** Current known models, sync. Static providers: the catalog. Dynamic providers: as of the last refresh (empty before the first). */
+  getModels(): readonly Model<TApi>[];
+
+  /** Dynamic providers only: fetch and update the model list. Concurrent calls share one in-flight fetch. */
+  refreshModels?(): Promise<void>;
+
+  stream<T extends TApi>(model: Model<T>, context: Context, options?: ApiStreamOptions<T>): AssistantMessageEventStream;
+
+  streamSimple(model: Model<TApi>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream;
+}
+```
+
+There is no `Provider.api` field. `model.api` carries API identity; the provider dispatches internally (see `createProvider()`).
+
+`Model.api` remains: existing metadata and tests use it, it is useful for diagnostics, and provider construction uses it for API implementation selection. But `Models` never dispatches on it; the provider does.
+
+### Typed stream options
+
+Full stream options are API-specific. `Model<TApi>` pays off by deriving the option type from the API:
+
+```ts
+// types.ts — type-only imports from API impl modules are erased, so this is tree-shake safe
+export interface ApiOptionsMap {
+  "anthropic-messages": AnthropicOptions;
+  "openai-completions": OpenAICompletionsOptions;
+  "openai-responses": OpenAIResponsesOptions;
+  "openai-codex-responses": OpenAICodexResponsesOptions;
+  "azure-openai-responses": AzureOpenAIResponsesOptions;
+  "google-generative-ai": GoogleOptions;
+  "google-vertex": GoogleVertexOptions;
+  "mistral-conversations": MistralOptions;
+  "bedrock-converse-stream": BedrockOptions;
+}
+
+export type ApiStreamOptions<TApi extends Api> = TApi extends keyof ApiOptionsMap
+  ? ApiOptionsMap[TApi]
+  : StreamOptions & Record<string, unknown>;
+```
+
+Custom api strings fall back to the generic shape.
+
+### Typed model narrowing
+
+Runtime model lists are dynamic, so `models.getModel()`/`getModels()` honestly return `Model<Api>`. Typing improves at three points:
+
+1. **`hasApi()` type guard** — runtime-checked narrowing for dynamic lookups (no blind casts):
+
+   ```ts
+   export function hasApi<TApi extends Api>(model: Model<Api>, api: TApi): model is Model<TApi>;
+
+   const model = await models.getModel("anthropic", "claude-opus-4-7");
+   if (model && hasApi(model, "anthropic-messages")) {
+     // model: Model<"anthropic-messages">, stream options fully typed
+   }
+   ```
+
+2. **`getBuiltinModel()`** — sync, generated-catalog lookup with typed overloads: `(provider, id) -> Model<exact-api-literal>`. The path for hardcoded known models.
+
+3. **`Provider<TApi>` factories** — typed model lists when using a provider directly, without a `Models` collection.
+
+Deliberately not done: tying `models.getModel(provider, ...)` to typed provider/model ids would require statically knowing which providers are installed in a mutable runtime collection. The harness path (`streamSimple` + `SimpleStreamOptions`) is API-agnostic and unaffected.
+
+For comparison: Vercel AI SDK attaches the implementation to the model object, which dissolves dispatch typing but makes models non-serializable (no sessions/RPC/catalogs as plain data), and its `providerOptions` bag is `Record<string, JSON>` checked only by `satisfies` convention. Plain-data models + provider-owned behavior keeps stronger typing where it matters.
+
+### Name collision
+
+`types.ts` currently exports `type Provider = KnownProvider | string` (a provider id). Rename that alias to `ProviderId` and fix call sites. The `Provider` interface above takes the name.
+
+## Provider model listing
+
+Reads are sync; fetching is an explicit async verb. `Provider.getModels()` returns the current known list — the full catalog for static providers, the last-refreshed list for dynamic ones (llama.cpp, OpenRouter live listing). `refreshModels()` is where dynamic providers fetch.
+
+This split exists because a sync-or-async union (`Promise<T> | T`) invites latent sync assumptions that detonate on the first async provider, while async-only reads force every consumer (UI lists, extension `find`/`getAll` surfaces) through Promises for data that is almost always static. Sync reads + explicit refresh keeps the staleness visible and the contract single: `getModels()` = last known, `refresh()` = make it current. A fetched list is stale the moment it returns anyway; naming the refresh point is honest about it.
+
+Apps own the refresh lifecycle: startup, registry reload, opening a model selector. Freshness-critical lookups are two-step: `await models.refresh("llamacpp"); models.getModel("llamacpp", id)`.
+
+Dynamic refresh must be side-effect-free discovery:
+
+```txt
+OK: fetch /v1/models, enumerate local catalog, refresh cached remote model list
+Not OK: load model, download model, mutate server state, run request probe
+```
+
+Provider-specific model lifecycle (load/unload) belongs in app/provider-management commands, not in `refreshModels()`.
+
+## Streaming path
+
+`Models.stream()` finds the provider by `model.provider`, resolves auth, merges it into request options, and delegates:
+
+```ts
+function stream(model, context, options) {
+  const provider = this.getProvider(model.provider);
+  if (!provider) {
+    // produce an error stream, not a throw — see Error behavior
+  }
+
+  // async setup happens inside the returned stream (lazyStream pattern)
+  const resolution = await this.getAuth(model);
+  const requestModel = resolution?.auth.baseUrl ? { ...model, baseUrl: resolution.auth.baseUrl } : model;
+  const requestOptions = mergeAuth(options, resolution?.auth); // explicit options win per-field
+
+  return provider.stream(requestModel, context, requestOptions);
+}
+```
+
+`stream()` returns `AssistantMessageEventStream` synchronously; async setup (auth resolution, lazy module load) happens inside the returned stream. The forwarding pattern already exists in today's `register-builtins.ts` (`createLazyStream`); extract it as `lazyStream()` in `src/api/lazy.ts`.
+
+No request hot-path model canonicalization: `stream()` uses the supplied model object as-is. If an app wants fresh model metadata, it refreshes the provider and re-reads (`await models.refresh(p); models.getModel(p, id)`) before starting the turn.
+
+## API implementations under `src/api`
+
+An API implementation is reusable stream behavior. It is not a provider.
+
+Uniform export contract — every real implementation module exports exactly:
+
+```ts
+// src/api/anthropic-messages.ts — imports SDKs
+export function stream(model, context, options) { ... }
+export function streamSimple(model, context, options) { ... }
+```
+
+This makes the module itself satisfy `ProviderStreams`, so the lazy wrapper is one generic helper instead of bespoke per-API plumbing. `ProviderStreams` is the untyped dispatch shape (implementation modules export concretely typed functions, which would not be assignable to a generic method); per-API option typing lives on the modules themselves and on `Provider.stream()` via `ApiStreamOptions`:
+
+```ts
+export interface ProviderStreams {
+  stream(model: Model<Api>, context: Context, options?: StreamOptions): AssistantMessageEventStream;
+  streamSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream;
+}
+
+// src/api/lazy.ts
+export function lazyApi(load: () => Promise<ProviderStreams>): ProviderStreams;
+
+// src/api/anthropic-messages.lazy.ts
+export const anthropicMessagesApi = (): ProviderStreams => lazyApi(() => import("./anthropic-messages.ts"));
+```
+
+Import chain:
+
+```txt
+provider module -> lazy API wrapper -> dynamic import(real API impl) -> SDK deps
+```
+
+Notes:
+
+- Bedrock keeps the node-only dynamic import trick (`importNodeOnlyProvider`, `.ts`/`.js` specifier rewrite) inside its lazy wrapper. `setBedrockProviderModule()` (used by the Bun build) moves into the bedrock lazy wrapper module.
+- Shared helper modules (`openai-responses-shared.ts`, `google-shared.ts`, `transform-messages.ts`, prompt-cache, copilot headers) move to `src/api/` alongside the implementations.
+
+## Shared API implementations across concrete providers
+
+Many concrete providers share an API implementation (OpenAI-completions: OpenRouter, Groq, Cerebras, xAI, ZAI, ...). They share lazy API objects by reference:
+
+```ts
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+
+export function openrouterProvider(): Provider {
+  return createProvider({
+    id: "openrouter",
+    name: "OpenRouter",
+    baseUrl: "https://openrouter.ai/api/v1",
+    auth: { apiKey: envApiKeyAuth("OpenRouter API key", ["OPENROUTER_API_KEY"]) },
+    models: OPENROUTER_MODELS,
+    api: openAICompletionsApi(),
+  });
+}
+```
+
+This copies Vercel AI SDK's useful property: users import concrete providers; shared protocol implementation is internal.
+
+## Auth
+
+Request auth output stays small:
+
+```ts
+export interface ModelAuth {
+  apiKey?: string;
+  headers?: Record<string, string>;
+  baseUrl?: string;
+}
+```
+
+If a value cannot be expressed as `apiKey`, `headers`, or `baseUrl`, it is provider config, not auth (Vertex project/location, Bedrock region/profile, Azure apiVersion are provider factory options).
+
+### Provider auth
+
+`Provider.auth` has exactly two slots; real providers have at most one api-key path and at most one OAuth path, and the slot names carry the UI's oauth-vs-api-key split without a `kind` discriminant or method ids:
+
+```ts
+export interface ProviderAuth {
+  apiKey?: ApiKeyAuth; // stored key/metadata + ambient env/files/ADC/IAM
+  oauth?: OAuthAuth;   // login flow + refresh
+}
+
+export interface ApiKeyAuth {
+  name: string; // "Anthropic API key"
+
+  /** Interactive setup (prompt for key/metadata). Absent = ambient-only (env, ADC, IAM). */
+  login?(callbacks: AuthLoginCallbacks): Promise<ApiKeyCredential>;
+
+  /**
+   * Resolve auth from the stored credential and/or ambient sources, merging
+   * per field (credential.key ?? env("..."), metadata.accountId ?? env("...")).
+   * undefined = not configured.
+   */
+  resolve(input: {
+    model: Model<Api>;
+    ctx: AuthContext;
+    credential?: ApiKeyCredential;
+  }): Promise<AuthResult | undefined>;
+}
+
+export interface OAuthAuth {
+  name: string; // "Anthropic (Claude Pro/Max)"
+
+  login(callbacks: AuthLoginCallbacks): Promise<OAuthCredential>;
+
+  /** Exchange the refresh token. Network call; throws on failure (invalid_grant etc.). Runs under the store lock. */
+  refresh(credential: OAuthCredential): Promise<OAuthCredential>;
+
+  /** Side-effect-free derivation of request auth from a valid credential. Covers Copilot-style per-credential baseUrl. Async so lazy wrappers can load the implementation. */
+  toAuth(credential: OAuthCredential): Promise<ModelAuth>;
+}
+
+export interface AuthResult {
+  auth: ModelAuth;
+  /** Human-readable label for status UI: "ANTHROPIC_API_KEY", "OAuth", "~/.aws/credentials". */
+  source?: string;
+}
+
+export interface AuthContext {
+  env(name: string): Promise<string | undefined>;
+  fileExists(path: string): Promise<boolean>; // supports leading ~
+}
+```
+
+The OAuth split (`refresh` + `toAuth` instead of one `resolve`) matches the old `OAuthProviderInterface` (`refreshToken` + `getApiKey`) and lets `Models` own the locking pattern without closure gymnastics: refresh produces a credential, `toAuth` derives request auth from whatever credential ends up stored.
+
+There is no `usesCallbackServer` flag. With `prompt()/notify()` callbacks the flow self-describes at runtime: a flow that runs a callback server issues a `manual_code` prompt racing the server and aborts the prompt when the callback wins. The UI needs no static foreknowledge.
+
+### Credentials
+
+One credential per provider, type-tagged — exactly the shape of today's auth.json (`type: "api_key" | "oauth"` per provider id):
+
+```ts
+export interface ApiKeyCredential {
+  type: "api-key";
+  key?: string;
+  metadata?: Record<string, string>; // e.g. Cloudflare accountId/gatewayId
+}
+
+export interface OAuthCredential extends OAuthCredentials {
+  type: "oauth"; // access, refresh, expires from OAuthCredentials
+}
+
+export type Credential = ApiKeyCredential | OAuthCredential;
+```
+
+`ApiKeyCredential.metadata` exists for providers like Cloudflare that store non-key values (account id, gateway id) alongside or instead of a key. `ApiKeyAuth.resolve()` merges per field: `credential.key ?? env("CLOUDFLARE_API_TOKEN")`, `credential.metadata?.accountId ?? env("CLOUDFLARE_ACCOUNT_ID")`, etc.
+
+### Credential store
+
+The app injects storage; `pi-ai` ships an in-memory default. Keyed by provider id, one credential per provider:
+
+```ts
+export interface CredentialStore {
+  /** Read the stored credential, possibly expired. Display/status use; request auth comes from Models.getAuth(). */
+  read(providerId: string): Promise<Credential | undefined>;
+
+  /**
+   * Serialized write — the only write path. fn sees the current credential
+   * because correct writes (refresh, login-during-refresh) depend on it;
+   * return the new credential, or undefined to leave the entry unchanged.
+   * Mutual exclusion per provider id, cross-process too where the backing
+   * store supports it (file lock). Resolves with the post-write credential.
+   */
+  modify(
+    providerId: string,
+    fn: (current: Credential | undefined) => Promise<Credential | undefined>,
+  ): Promise<Credential | undefined>;
+
+  /** Remove (logout). Serialized against modify. */
+  delete(providerId: string): Promise<void>;
+}
+```
+
+There is deliberately no `set`: an unserialized write path invites read-modify-write races (login-during-refresh clobbering a fresh credential, double token refresh). Call sites:
+
+```ts
+await store.modify(pid, async () => credential);      // login: store this
+await store.read(pid);                                // status UI ("logged in via OAuth")
+await store.delete(pid);                               // logout
+// refresh RMW happens inside Models.getAuth
+```
+
+Error semantics: `read` resolves `undefined` for missing entries; methods reject only on storage failure, and `Models` wraps such rejections in `ModelsError` code `"auth"`. Best-effort stores that serve an in-memory view and record persistence errors internally (today's AuthStorage behavior) are valid implementations.
+
+### Resolution policy (fixed)
+
+`Models.getAuth(model)` is a decision tree, not a loop. A stored credential owns the provider — ambient/env is consulted only when nothing is stored (AuthStorage parity: no silent env fallback after a failed refresh or for an unmatched credential type):
+
+```ts
+const stored = await store.read(provider.id);
+if (stored) {
+  if (stored.type === "oauth" && provider.auth.oauth) {
+    const oauth = provider.auth.oauth;
+    let credential = stored;
+    if (Date.now() >= credential.expires) {                 // optimistic check, lock-free
+      const post = await store.modify(provider.id, async (current) => {
+        if (current?.type !== "oauth") return undefined;    // logged out meanwhile
+        return Date.now() >= current.expires                // authoritative check, under lock
+          ? oauth.refresh(current)                          // throws -> ModelsError("oauth")
+          : undefined;                                      // another process/request refreshed
+      });
+      if (post?.type !== "oauth") return undefined;
+      credential = post;
+    }
+    return { auth: await oauth.toAuth(credential), source: "OAuth" };
+  }
+  if (stored.type === "api-key" && provider.auth.apiKey) {
+    return provider.auth.apiKey.resolve({ model, ctx, credential: stored });
+  }
+  return undefined; // stored credential without matching handler blocks ambient
+}
+return provider.auth.apiKey?.resolve({ model, ctx, credential: undefined }); // ambient
+```
+
+Properties:
+
+- Double-checked locking, same as today's `refreshOAuthTokenWithLock`: valid tokens cost one `read` and zero locks; expired tokens lock, re-check under the lock, refresh once globally, persist before release.
+- Explicit request auth (stream options `apiKey`/`headers`) is merged per-field on top in `stream()`, winning over everything.
+- Refresh failure rejects with `ModelsError("oauth")`; the stored credential is untouched (preserved for retry). Request paths surface this as a stream error with the real cause ("run /login"); status/availability UIs catch the rejection and render "needs re-login" — documented contract on `getAuth`.
+
+### Replacing AuthStorage
+
+The end state for coding-agent: AuthStorage is deleted; its capabilities map onto a `CredentialStore` implementation plus composition.
+
+Today's `getApiKey` priority and its new home:
+
+| AuthStorage today | New design |
+|---|---|
+| runtime override (CLI `--api-key`) | `withRuntimeOverrides(store, overrides)` decorator: `read` returns the override as an `ApiKeyCredential`; never persisted |
+| stored `api_key` (with `$ENV`/`!command` via `resolveConfigValue`) | stored `ApiKeyCredential`; config-value resolution happens at `read` in coding-agent's adapter/decorator (command execution stays app policy) |
+| stored `oauth` + locked refresh, undefined on failure | `getAuth` decision tree above; failure rejects with cause instead of silently unconfiguring |
+| env var (only when nothing stored) | ambient branch of `apiKey.resolve` |
+| `fallbackResolver` (models.json custom providers) | gone — custom providers carry their own `auth.apiKey` |
+
+```txt
+FileCredentialStore        ports AuthStorage's lock backend: read = memory snapshot,
+                           modify = withLockAsync(re-read, fn, merge-write), delete,
+                           internal error recording (drainErrors equivalent)
+└─ withConfigValues        $ENV / !command at read
+   └─ withRuntimeOverrides --api-key
+      └─ createModels({ credentials: store })
+
+login/logout UI            provider.auth.{oauth,apiKey}.login(callbacks) + store.modify/delete
+status UI                  store.read(pid) + getAuth try/catch ("needs /login" on rejection)
+getOAuthProviders          presence of provider.auth.oauth across registered providers
+```
+
+### Login callbacks
+
+One interface serves api-key and OAuth login:
+
+```ts
+export interface AuthLoginCallbacks {
+  /** Aborts the whole login flow. Per-prompt cancellation uses AuthPrompt.signal. */
+  signal?: AbortSignal;
+
+  prompt(prompt: AuthPrompt): Promise<string>;
+  notify(event: AuthEvent): void;
+}
+
+/** `signal` lets the flow cancel a pending prompt when an out-of-band event resolves the step. */
+export type AuthPrompt = { signal?: AbortSignal } & (
+  | { type: "text"; message: string; placeholder?: string }
+  | { type: "secret"; message: string; placeholder?: string }
+  | { type: "select"; message: string; options: readonly { id: string; label: string; description?: string }[] }
+  | { type: "manual_code"; message: string; placeholder?: string }
+);
+
+export type AuthEvent =
+  | { type: "auth_url"; url: string; instructions?: string }
+  | { type: "device_code"; userCode: string; verificationUri: string; intervalSeconds?: number; expiresInSeconds?: number }
+  | { type: "progress"; message: string };
+```
+
+`prompt()` returns the entered/selected string (`select` returns the option id). Flows race a `manual_code` prompt against a callback server by setting `AuthPrompt.signal` and aborting the prompt when the callback wins.
+
+### OAuth attachment
+
+Providers that support OAuth always attach it. There is no factory toggle: the flow is lazy-loaded, so advertising OAuth costs nothing until `login()`/`refresh()` actually runs, and a host that never logs in never loads it.
+
+```ts
+export function anthropicProvider(): Provider {
+  return createProvider({
+    id: "anthropic",
+    name: "Anthropic",
+    baseUrl: "https://api.anthropic.com/v1",
+    auth: {
+      apiKey: envApiKeyAuth("Anthropic API key", ["ANTHROPIC_API_KEY"]),
+      oauth: lazyOAuth({
+        name: "Anthropic (Claude Pro/Max)",
+        load: () => import("../utils/oauth/anthropic.ts").then((m) => m.anthropicOAuth),
+      }),
+    },
+    models: ANTHROPIC_MODELS,
+    api: anthropicMessagesApi(),
+  });
+}
+```
+
+`lazyOAuth()` wraps a dynamically imported `OAuthAuth` so provider definitions can advertise OAuth without importing the implementation (`toAuth` is async for exactly this reason):
+
+```ts
+export function lazyOAuth(input: {
+  name: string;
+  load: () => Promise<OAuthAuth>;
+}): OAuthAuth;
+```
+
+OAuth must not force Node-only code (`node:http`, `node:crypto`) into browser bundles: the dynamic import inside `lazyOAuth()` uses the same bundler-opaque variable-specifier trick as the bedrock lazy wrapper. Browser hosts never trigger the load (no stored node OAuth credentials, no login flow). If web OAuth lands later (sitegeist proved feasibility: Web Crypto PKCE, auth tab, fetch token exchange, device-code polling), it is just a different `OAuthAuth` implementation — no reserved option values.
+
+The existing flows in `src/utils/oauth/` (anthropic, openai-codex, github-copilot) are adapted to `OAuthAuth` (`login`/`refresh`/`toAuth`, replacing `login`/`refreshToken`/`getApiKey`/`modifyModels`) with the new callbacks, staying Node-targeted and lazy-loaded. Copilot's `modifyModels` baseUrl rewriting becomes `toAuth` returning `ModelAuth.baseUrl`.
+
+## Provider wrappers and models.json
+
+`models.json` is a provider wrapper layer. It does not mutate providers in place:
+
+```ts
+function withProviderOverrides(base: Provider, overrides: ProviderOverrides): Provider {
+  return {
+    ...base,
+    name: overrides.name ?? base.name,
+    baseUrl: overrides.baseUrl ?? base.baseUrl,
+    headers: mergeHeaders(base.headers, overrides.headers),
+
+    getModels: () => applyModelOverrides(base.getModels(), overrides.models),
+    refreshModels: base.refreshModels?.bind(base),
+
+    stream: base.stream,
+    streamSimple: base.streamSimple,
+  };
+}
+```
+
+This composes with dynamic providers because `getModels()` delegates to the base source and `refreshModels()` passes through.
+
+Request-auth config from models.json (`$ENV`, `!command`, inline keys) remains app-owned sidecar state, surfaced either as explicit request auth or as a custom `ApiKeyAuth` the app sets on the wrapped provider's `auth.apiKey`.
+
+## Custom providers: createProvider()
+
+One helper builds providers from parts; it handles both single-API and mixed-API providers:
+
+```ts
+export function createProvider(input: {
+  id: string;
+  name?: string;                 // default: id
+  baseUrl?: string;
+  headers?: Record<string, string>;
+  auth: ProviderAuth;            // required, at least one of apiKey/oauth (no "no-auth" providers)
+  /** Initial model list (empty for purely dynamic providers). */
+  models: readonly Model<Api>[];
+  /** Dynamic providers: fetch the current list; createProvider stores it and dedupes in-flight calls. */
+  refreshModels?: () => Promise<readonly Model<Api>[]>;
+  /** Single implementation, or map keyed by model.api for mixed-API providers. */
+  api: ProviderStreams | Record<string, ProviderStreams>;
+}): Provider;
+```
+
+- Single `api`: all models stream through it.
+- Map `api`: `stream()`/`streamSimple()` dispatch on `model.api`; unknown api produces a stream error.
+
+Mixed-API custom providers must be supported (opencode Go/Zen-style providers expose models backed by different APIs under one provider id).
+
+Built-in provider factories use `createProvider()` internally. models.json custom providers map onto it directly:
+
+```json
+{
+  "providers": {
+    "my-openai-proxy": {
+      "api": "openai-completions",
+      "baseUrl": "https://proxy.example/v1",
+      "models": [ ... ]
+    }
+  }
+}
+```
+
+## Compat entrypoint
+
+`@earendil-works/pi-ai/compat` preserves the old global API surface until the coding-agent migration deletes it. New code never imports it.
+
+Old semantics being preserved: global `stream()` dispatched purely on `model.api` via the api-registry, with env API key injection. The compat module reproduces this exactly — it does not route through a `Models` collection, so compat consumers get zero behavioral drift (a `Models`-routed variant was considered and dropped: a model with a known provider id but a different api would dispatch wrong, and auth semantics would shift mid-migration). The harness `Models` instance (Phase 6/7) is where new-path streaming happens.
+
+- `stream/complete/streamSimple/completeSimple(model, ctx, opts)`: api-dispatch via the api-registry plus `getEnvApiKey` injection, verbatim old behavior.
+- The builtin api registration side effect moves from the root barrel into compat. It skips api ids that already have a registration, since compat may load after a test or extension registered an override. `registerApiProvider()/unregisterApiProviders()` keep feeding the registry; `resetApiProviders()` clears and re-registers builtins.
+- Sync `getModel/getModels/getProviders` are deprecated aliases of `getBuiltinModel/getBuiltinModels/getBuiltinProviders` from `providers/all` (they were always pure generated-catalog reads — verified: nothing ever mutated the old `modelRegistry`).
+- Re-exports the per-API lazy stream wrappers (incl. `setBedrockProviderModule`), `env-api-keys.ts`, and the image-generation registry/catalogs; none of these stay on the root barrel.
+- `export * from "./index.ts"`: compat is a strict superset of the core entrypoint, so consumers switch a file's import path wholesale without symbol surgery.
+
+coding-agent (and the interim agent package) switch imports of these symbols from `@earendil-works/pi-ai` to `@earendil-works/pi-ai/compat` (import-path-only change) and are otherwise untouched until the ModelManager migration.
+
+Extension grace period: the coding-agent extension loader (jiti aliases + Bun `virtualModules`) resolves the `@earendil-works/pi-ai` ROOT specifier to the compat entrypoint. Existing user extensions using the old global API (`complete`, `getModel`, `registerApiProvider`, ...) keep working at runtime without changes; they break only when compat is removed at the ModelManager migration, with a migration guide in the changelog. Typechecking is the nudge: editors resolve the root to the slim core types, so extension sources that typecheck must import old globals from `/compat` — which is what the repo example extensions demonstrate.
+
+## Builtin static helpers
+
+Typed, sync, generated-catalog-only helpers live with the catalogs (exported from `providers/all`):
+
+```ts
+getBuiltinModel(provider, id)   // sync, typed overloads from generated catalog
+getBuiltinModels(provider)      // sync
+getBuiltinProviders()           // sync
+```
+
+Runtime lookup is always the async instance API: `await models.getModel(...)`.
+
+Generated catalogs are split per provider (`providers/<id>.models.ts`) by updating `packages/ai/scripts/generate-models.ts`. If the generator change turns out too large for this pass, splitting may be deferred; `providers/all` and provider factories may temporarily import the monolithic `models.generated.ts`, relying on `sideEffects: false` for pruning.
+
+## Tree-shaking and lazy imports
+
+Rules:
+
+1. Main `@earendil-works/pi-ai` import is core-only.
+2. Provider modules import their catalog, auth helpers, and lazy API wrappers only.
+3. Lazy API wrappers dynamically import real API implementations.
+4. Real API implementations import SDK dependencies.
+5. OAuth implementations are always attached via `lazyOAuth()` and lazy-loaded behind a bundler-opaque dynamic import; provider metadata never eagerly imports Node-only OAuth code.
+6. `providers/all` may import all provider metadata, but no eager SDK imports.
+7. Provider modules are side-effect-free; importing a provider does not register anything globally.
+8. `package.json` sets `sideEffects: false`.
+
+Exports map sketch:
+
+```json
+{
+  "exports": {
+    ".": "./dist/index.js",
+    "./compat": "./dist/compat.js",
+    "./providers/all": "./dist/providers/all.js",
+    "./providers/openai": "./dist/providers/openai.js",
+    "./providers/anthropic": "./dist/providers/anthropic.js",
+    "./providers/*": "./dist/providers/*.js",
+    "./api/*": "./dist/api/*.js"
+  }
+}
+```
+
+Browser smoke check (`scripts/check-browser-smoke.mjs`) must keep passing: bundling the core entrypoint (and any non-node provider entrypoint) must not pull `node:http`/`node:crypto`.
+
+## AgentHarness integration
+
+`AgentHarness` receives a `Models` instance.
+
+- `AgentHarnessOptions.models` is required.
+- The harness does not snapshot `Models` into turn state.
+- Request path calls `this.models.streamSimple(model, context, options)`; same for compaction/branch-summarization paths.
+- Request path never calls async `models.getModel()` to canonicalize; if model metadata needs refresh, the app updates the selected model before starting a turn.
+- Harness tests build `createModels()` and install the faux provider (`fauxProvider()` factory from `providers/faux`).
+
+## coding-agent next phase (not this pass)
+
+coding-agent builds providers in layers and binds them per session:
+
+```txt
+built-in providers (builtinModels)
+-> models.json provider wrappers / custom providers (createProvider)
+-> extension provider wrappers/additions
+```
+
+```ts
+sessionModels.clearProviders();
+for (const provider of layeredProviders) sessionModels.setProvider(provider);
+```
+
+coding-agent owns: `FileCredentialStore` + decorators replacing AuthStorage (see "Replacing AuthStorage"), models.json auth sidecar (`$ENV`, `!command`), command execution policy, provider status labels (from `AuthResult.source`), login/logout UI (driving `auth.{apiKey,oauth}.login()` with `prompt()/notify()`), extension lifecycle, provider-management slash commands.
+
+Until then, the only coding-agent changes in this pass are:
+
+- construct a `Models` instance for `AgentHarness` (builtins + legacy api-dispatch fallback bridging `ModelRegistry` custom providers)
+- switch old-global imports to `@earendil-works/pi-ai/compat`
+- adapt the login dialog to `prompt()/notify()` callbacks (thin UI adapter; replaces the `usesCallbackServer` special-casing)
+
+## Implementation TODOs
+
+Check items off as they land. Keep this list current; it is the working state for resumed sessions.
+
+### Phase 1 — core types/runtime
+
+- [x] Rename `types.ts` `Provider` alias to `ProviderId`; fix call sites.
+- [x] Add `ApiOptionsMap` and `ApiStreamOptions<TApi>` to `types.ts` (type-only imports).
+- [x] New `models.ts`: `Provider<TApi>` interface, `hasApi()` guard, `ModelsError` + codes. Auth types live in `src/auth/types.ts` (`ProviderAuth` = `{ apiKey?, oauth? }`, credentials, `CredentialStore` (`read`/`modify`/`delete`, one credential per provider), `AuthResult`, `AuthContext`, `ModelAuth`, login callbacks), in-memory store in `src/auth/credential-store.ts`, default context in `src/auth/context.ts` (browser-safe node:fs trick), `lazyStream()` in `src/api/lazy.ts`.
+- [x] `Models`/`MutableModels`/`createModels({ credentials?, authContext? })` with provider map, async `getModel(s)` (per-provider failure isolation), `getAuth` (decision tree, double-checked locked refresh), `stream/complete/streamSimple/completeSimple` with per-field auth merge. Tests: `packages/ai/test/models-runtime.test.ts`.
+- [x] Keep metadata helpers: `calculateCost`, `getSupportedThinkingLevels`, `clampThinkingLevel`, `modelsAreEqual`.
+
+### Phase 2 — `src/api/`
+
+- [x] Move stream implementations from `src/providers/` to `src/api/`, renamed by API id (`anthropic.ts` -> `api/anthropic-messages.ts`, etc.).
+- [x] Normalize each implementation module to export exactly `stream` and `streamSimple`.
+- [x] Move shared helpers (`openai-responses-shared`, `google-shared`, `transform-messages`, `openai-prompt-cache`, `github-copilot-headers`, `cloudflare`, `simple-options`) to `src/api/`.
+- [x] Extract `lazyStream()`/`lazyApi()` into `src/api/lazy.ts`.
+- [x] Add `*.lazy.ts` wrappers per API; bedrock keeps node-only import trick and `setBedrockProviderModule()`.
+- [x] Delete `providers/register-builtins.ts`. Interim until Phase 5 compat: builtin api-registry registration lives in `stream.ts`; lazy API wrappers are exported from the root barrel.
+
+### Phase 3 — provider factories + catalogs
+
+- [x] Auth helpers in `src/auth/helpers.ts`: `envApiKeyAuth()` (with secret-prompt `login`), `lazyOAuth()`. OAuth flow loads go through `utils/oauth/load.ts` (bundler-opaque dynamic import); the `OAuthAuth` exports it references land in Phase 4.
+- [x] `createProvider()` in `models.ts` (single + mixed `api` map, dispatch on `model.api`, unknown api -> stream error).
+- [x] Per-provider factories under `src/providers/` for all built-in catalog providers; OAuth attached via `lazyOAuth()` (anthropic, openai-codex, github-copilot); ambient `ApiKeyAuth` for amazon-bedrock (AWS env/profile) and google-vertex (key or ADC+project+location).
+- [x] `providers/all.ts`: `builtinProviders()`, `builtinModels()`, `getBuiltinModel/getBuiltinModels/getBuiltinProviders` re-exports.
+- [x] Faux provider factory (`fauxProvider()` in `providers/faux.ts`) for tests; legacy `registerFauxProvider()` kept until compat dies.
+- [x] Split generated catalogs per provider via `scripts/generate-models.ts` (`providers/<id>.models.ts`); `models.generated.ts` becomes a generated aggregator.
+
+### Phase 4 — OAuth adaptation
+
+- [x] Adapt `utils/oauth/anthropic.ts`, `openai-codex.ts`, `github-copilot.ts` to `OAuthAuth` (`login`/`refresh`/`toAuth`) + `prompt()/notify()`; `modifyModels` baseUrl rewriting becomes `toAuth().baseUrl`. New exports (`anthropicOAuth`, `openaiCodexOAuth`, `githubCopilotOAuth`) sit next to the old `OAuthProviderInterface` objects, which survive until Phase 7.
+- [x] No `usesCallbackServer` on `OAuthAuth`: callback-server flows race a `manual_code` prompt (aborted via `AuthPrompt.signal` once the flow settles). The old interface keeps its flag until it dies with compat.
+
+### Phase 5 — packaging
+
+- [x] `index.ts` core-only and side-effect free (no catalogs, no provider factories, no api-registry, no env-api-keys, no images, no OAuth, no compat). Typed catalog reads (`getBuiltin*`) implemented in `providers/all.ts`; `models.ts` no longer imports `models.generated.ts`.
+- [x] `compat.ts`: superset of index + old api-dispatch globals, deprecated `getModel/getModels/getProviders` aliases, lazy api wrappers + `setBedrockProviderModule`, `getEnvApiKey`, images. Registration side effect lives here (skip-if-present).
+- [x] Subpath exports map (`./compat`, `./providers/*`, `./api/*`); `sideEffects` array listing the effectful modules (`compat`, images registration) instead of `false`.
+- [x] Browser smoke (entry now imports old globals from `/compat`) + shrinkwrap checks green. Internal old-global imports switched to `/compat` already (42 files in agent/coding-agent/examples; vitest configs alias `/compat` to src; spawn-CLI tests resolve workspace dist, so `packages/ai` + `packages/agent` dists were rebuilt).
+
+### Phase 6 — AgentHarness
+
+- [x] `AgentHarnessOptions.models` required (`readonly models` on the harness); the harness stream path uses `models.streamSimple()`. `StreamFn` redefined structurally (no compat type dependency); `Models.streamSimple` satisfies it.
+- [x] Compaction/branch-summarization take the harness `Models` instance. `getApiKeyAndHeaders` is removed entirely — `Models` is the only auth path; per-request key resolution becomes provider auth on the collection. `compact()`/`generateSummary()`/`generateBranchSummary()` lose their explicit `apiKey`/`headers` parameters.
+- [x] Harness tests use `createModels()` + `fauxProvider()` with unique per-fake provider ids; no global api-registry state, no unregister bookkeeping.
+
+### Phase 7 — coding-agent bridge (minimal)
+
+- [x] Switch old-global imports to `@earendil-works/pi-ai/compat` (landed with Phase 5; compat is a superset so the switch was path-only). Extension loader resolves the pi-ai root to compat as the runtime grace period.
+- [x] Everything else originally sketched here is gated on coding-agent actually streaming through a `Models` instance — coding-agent's `AgentSession` drives the low-level `Agent` via `streamFn`, not the harness — and moved to Phase 9.
+
+### Phase 8 — wrap-up
+
+- [x] Update/add tests; run affected suites (tests landed with each phase; `./test.sh` green throughout).
+- [x] `packages/ai/CHANGELOG.md`: `### Breaking Changes` with migration guide (compat entrypoint, `Provider` -> `ProviderId`, api module moves) + `### Added` for the new Models/provider/auth API.
+- [x] `packages/coding-agent/CHANGELOG.md`: `### Changed` entry for extension authors — runtime unaffected (loader resolves the pi-ai root to compat), typecheck nudges to `/compat` or the new API; removal happens later with a migration guide.
+- [x] `packages/agent/CHANGELOG.md`: `### Breaking Changes` for required `AgentHarnessOptions.models`, compaction signature changes, structural `StreamFn`.
+- [x] `npm run check` clean.
+
+### Phase 9 — coding-agent on Models + CredentialStore (in scope)
+
+coding-agent replaces AuthStorage and ModelRegistry's internals with `FileCredentialStore` + a `MutableModels` collection. AgentSession itself stays (AgentHarness adoption is pi 2.0); only its model/auth substrate swaps. Layering is strictly one-directional:
+
+```txt
+FileCredentialStore (auth.json, locked) + --api-key overlay + $ENV/!command resolution
+        ↑
+MutableModels: builtin factories (wrapped per models.json config) + custom providers (models.json ∪ extensions)
+        ↑
+ModelRegistry: async facade — reads delegate to the collection; registerProvider/login/logout/status for extensions + UI
+        ↑
+AgentSession / sdk / interactive-mode (await added; stream via models)
+```
+
+Decisions:
+
+- `AuthStorage` is deleted as a type — it would otherwise depend on provider auth while provider auth depends on its store (circular). Its surface splits: `get`/`set`/`remove` -> `CredentialStore`; `getApiKey` -> `Models.getAuth`; `login`/`logout`/`getAuthStatus` -> ModelRegistry facade methods over `provider.auth.oauth` + the store.
+- Runtime `--api-key` overrides are a store overlay (an override reads as an ephemeral stored api-key credential, masking stored OAuth — matches today's priority). Every registered provider is guaranteed an `apiKey` auth slot so overrides apply to OAuth-only providers too.
+- `ModelRegistry.getAll`/`find`/`getAvailable` become async, delegating to the collection (no materialized snapshot, no sync lies; dynamic providers like llama.cpp work). The extension-facing `modelRegistry` surface changes accordingly (breaking, changelogged); extensions also get the collection itself as the forward API.
+- models.json keeps FULL feature parity, implemented as provider decoration: builtin factories wrapped so `getModels()` applies provider `baseUrl`/`compat` overlays, `modelOverrides`, and custom-model merges (async-safe); provider `apiKey`/`headers`/`authHeader` configs become that provider's `ApiKeyAuth` (config first, factory auth fallback); parse errors keep `getError()` semantics.
+- Extension `ProviderConfig` parity: provider-keyed `streamSimple`, old-style `oauth` adapted to `OAuthAuth` (`modifyModels` -> `getModels` wrap + `toAuth`), full model replacement per provider. The api-registry `registerApiProvider` dual-write stays for compat consumers (extensions calling global `complete()`); it dies with compat.
+- Copilot: stored-credential baseUrl applied in the wrapped `getModels()` (extension-visible models stay correct) plus per-request `toAuth().baseUrl`.
+- Cloudflare: provider-auth substitution (key + `CLOUDFLARE_ACCOUNT_ID`/`CLOUDFLARE_GATEWAY_ID` from credential metadata/env -> `ModelAuth.baseUrl`); impl-side `resolveCloudflareBaseUrl` stays until compat dies (it is idempotent on substituted URLs), keeping extension `complete()` calls working.
+
+Ordering:
+
+- [ ] pi-ai rework first: `Provider.getModels()` sync + optional `refreshModels()`; `Models.getModels`/`getModel` sync, `Models.refresh(provider?)` async; `createProvider` takes `models` array + optional `refreshModels` fetcher (in-flight dedupe). Reverses Phase 1's async-listing decision — see "Provider model listing" for rationale (sync-or-async unions breed latent sync assumptions; async-only breaks sync consumer surfaces like extension `find`/`getAll`).
+- [ ] `FileCredentialStore` (ports the auth.json lock backend, reads legacy `type: "api_key"` tags) + `--api-key` overlay + `$ENV`/`!command` resolution; tests.
+- [ ] Cloudflare provider auth in pi-ai factories; copilot `getModels` baseUrl wrap.
+- [ ] Extension-OAuth adapter (old `OAuthProviderInterface` config -> `OAuthAuth`).
+- [ ] ModelRegistry rebuild: owns `MutableModels`, async reads, models.json decoration, provider-keyed extension streams, facade auth ops; AuthStorage deleted.
+- [ ] Consumer rewiring: agent-session, sdk (`credentials?: CredentialStore` option replaces `authStorage`; sdk.md updated), model-resolver, interactive login/status UI on `prompt()/notify()`, cli `--api-key`.
+- [ ] Test migration; tmux validation of login flows against real providers.
+
+### Phase 10 — compat deletion (pi 2.0 era, separate)
+
+- [ ] AgentSession -> AgentHarness; the registry facade dies in favor of harness `Models`.
+- [ ] Move ALL internal `/compat` imports to the new API: every package's src, all tests, and the example extensions (examples then demonstrate the new API). Nothing inside the repo may import `/compat` at that point.
+- [ ] Delete `/compat`, `api-registry.ts`, `env-api-keys.ts`, the extension-loader root-to-compat alias, the old `pi-ai/oauth` registry and `OAuthProviderInterface` (incl. `usesCallbackServer`), and the impl-side cloudflare substitution. This is the extension-author breaking release; changelog carries the migration guide.
+
+### Deferred / follow-ups
+
+- [ ] Web OAuth implementations (sitegeist-style) as an alternative `OAuthAuth`.
+- [x] Images API redesign: `ImagesModels`/`ImagesProvider`/`createImagesProvider` mirror the chat-side design (sync reads, explicit refresh, never-reject generation); auth resolution shared with the chat side via the free-standing `resolveProviderAuth()` in `auth/resolve.ts` (which also owns `ModelsError`; both collections pass their store/context as arguments — no resolver object). `openrouterImagesProvider()` factory + `builtinImagesProviders()`/`builtinImagesModels()` in `providers/all`; impl moved to `api/openrouter-images.ts` with a lazy wrapper. The old global image API (registry + `getImageModel*` + `generateImages`) stays on compat; `ImagesProvider` id alias in types.ts renamed to `ImagesProviderId` (mirror of `Provider` -> `ProviderId`).
+
+## Error behavior
+
+`undefined` means not found or not configured. Real failures reject or become stream errors.
+
+```ts
+export type ModelsErrorCode =
+  | "model_source"      // provider model refresh failed
+  | "model_validation"  // model object invalid
+  | "provider"          // unknown provider, dispatch failure
+  | "stream"            // stream setup failure
+  | "auth"              // auth resolution failure
+  | "oauth";            // oauth login/refresh failure
+```
+
+- `Models.stream()` produces stream errors (error event + error result) for async setup failures; it does not throw after returning the stream.
+- `Models.getModels()` is a sync best-effort read: a provider whose `getModels()` throws yields no models. `Models.refresh(provider)` rejects on that provider's fetch failure; `Models.refresh()` (all providers) is concurrent best-effort. Apps that need a concrete listing failure refresh the single provider.
+- Auth resolution and credential store failures reject loudly (`ModelsError` codes `auth`/`oauth`); silent fallback to a different auth path after a failure risks billing surprises. A stored credential always blocks ambient/env fallback, including after a failed refresh.
+- Status/availability UIs catch `getAuth` rejections and render "needs re-login"; they do not treat rejection as "unconfigured".

packages/agent/src/agent-loop.ts

@@ -10,7 +10,7 @@ import {
 	streamSimple,
 	type ToolResultMessage,
 	validateToolArguments,
-} from "@earendil-works/pi-ai";
+} from "@earendil-works/pi-ai/compat";
 import type {
 	AgentContext,
 	AgentEvent,

packages/agent/src/agent.ts

@@ -7,7 +7,7 @@ import {
 	type TextContent,
 	type ThinkingBudgets,
 	type Transport,
-} from "@earendil-works/pi-ai";
+} from "@earendil-works/pi-ai/compat";
 import { runAgentLoop, runAgentLoopContinue } from "./agent-loop.ts";
 import type {
 	AfterToolCallContext,

packages/agent/src/harness/agent-harness.ts

@@ -1,10 +1,4 @@
-import {
-	type AssistantMessage,
-	type ImageContent,
-	type Model,
-	streamSimple,
-	type UserMessage,
-} from "@earendil-works/pi-ai";
+import type { AssistantMessage, ImageContent, Model, Models, UserMessage } from "@earendil-works/pi-ai";
 import { runAgentLoop } from "../agent-loop.ts";
 import type {
 	AgentContext,
@@ -75,17 +69,6 @@ function cloneStreamOptions(streamOptions?: AgentHarnessStreamOptions): AgentHar
 	};
 }
 
-function mergeHeaders(...headers: Array<Record<string, string> | undefined>): Record<string, string> | undefined {
-	const merged: Record<string, string> = {};
-	let hasHeaders = false;
-	for (const entry of headers) {
-		if (!entry) continue;
-		Object.assign(merged, entry);
-		hasHeaders = true;
-	}
-	return hasHeaders ? merged : undefined;
-}
-
 function findDuplicateNames(names: string[]): string[] {
 	const seen = new Set<string>();
 	const duplicates = new Set<string>();
@@ -178,6 +161,7 @@ export class AgentHarness<
 > {
 	readonly env: ExecutionEnv;
 	private session: Session;
+	readonly models: Models;
 	private phase: AgentHarnessPhase = "idle";
 	private runAbortController?: AbortController;
 	private runPromise?: Promise<void>;
@@ -186,7 +170,6 @@ export class AgentHarness<
 	private thinkingLevel: ThinkingLevel;
 	private systemPrompt: AgentHarnessOptions<TSkill, TPromptTemplate, TTool>["systemPrompt"];
 	private streamOptions: AgentHarnessStreamOptions;
-	private getApiKeyAndHeaders?: AgentHarnessOptions["getApiKeyAndHeaders"];
 	private resources: AgentHarnessResources<TSkill, TPromptTemplate>;
 	private tools = new Map<string, TTool>();
 	private activeToolNames: string[];
@@ -200,10 +183,10 @@ export class AgentHarness<
 	constructor(options: AgentHarnessOptions<TSkill, TPromptTemplate, TTool>) {
 		this.env = options.env;
 		this.session = options.session;
+		this.models = options.models;
 		this.resources = options.resources ?? {};
 		this.streamOptions = cloneStreamOptions(options.streamOptions);
 		this.systemPrompt = options.systemPrompt;
-		this.getApiKeyAndHeaders = options.getApiKeyAndHeaders;
 		this.validateUniqueNames(
 			(options.tools ?? []).map((tool) => tool.name),
 			"Duplicate tool name(s)",
@@ -376,13 +359,9 @@ export class AgentHarness<
 	private createStreamFn(getTurnState: () => AgentHarnessTurnState<TSkill, TPromptTemplate, TTool>): StreamFn {
 		return async (model, context, streamOptions) => {
 			const turnState = getTurnState();
-			const auth = await this.getApiKeyAndHeaders?.(model);
-			const snapshotOptions: AgentHarnessStreamOptions = {
-				...turnState.streamOptions,
-				headers: mergeHeaders(turnState.streamOptions.headers, auth?.headers),
-			};
+			const snapshotOptions: AgentHarnessStreamOptions = { ...turnState.streamOptions };
 			const requestOptions = await this.emitBeforeProviderRequest(model, turnState.sessionId, snapshotOptions);
-			return streamSimple(model, context, {
+			return this.models.streamSimple(model, context, {
 				cacheRetention: requestOptions.cacheRetention,
 				headers: requestOptions.headers,
 				maxRetries: requestOptions.maxRetries,
@@ -401,7 +380,6 @@ export class AgentHarness<
 				sessionId: turnState.sessionId,
 				timeoutMs: requestOptions.timeoutMs,
 				transport: requestOptions.transport,
-				apiKey: auth?.apiKey,
 			});
 		};
 	}
@@ -713,8 +691,6 @@ export class AgentHarness<
 		try {
 			const model = this.model;
 			if (!model) throw new AgentHarnessError("invalid_state", "No model set for compaction");
-			const auth = await this.getApiKeyAndHeaders?.(model);
-			if (!auth) throw new AgentHarnessError("auth", "No auth available for compaction");
 			const branchEntries = await this.session.getBranch();
 			const preparationResult = prepareCompaction(branchEntries, DEFAULT_COMPACTION_SETTINGS);
 			if (!preparationResult.ok) throw preparationResult.error;
@@ -731,15 +707,7 @@ export class AgentHarness<
 			const provided = hookResult?.compaction;
 			const compactResult = provided
 				? { ok: true as const, value: provided }
-				: await compact(
-						preparation,
-						model,
-						auth.apiKey,
-						auth.headers,
-						customInstructions,
-						undefined,
-						this.thinkingLevel,
-					);
+				: await compact(preparation, this.models, model, customInstructions, undefined, this.thinkingLevel);
 			if (!compactResult.ok) throw compactResult.error;
 			const result = compactResult.value;
 			const entryId = await this.session.appendCompaction(
@@ -792,12 +760,9 @@ export class AgentHarness<
 			if (!summaryText && options?.summarize && entries.length > 0) {
 				const model = this.model;
 				if (!model) throw new AgentHarnessError("invalid_state", "No model set for branch summary");
-				const auth = await this.getApiKeyAndHeaders?.(model);
-				if (!auth) throw new AgentHarnessError("auth", "No auth available for branch summary");
 				const branchSummary = await generateBranchSummary(entries, {
+					models: this.models,
 					model,
-					apiKey: auth.apiKey,
-					headers: auth.headers,
 					signal: new AbortController().signal,
 					customInstructions: hookResult?.customInstructions ?? options?.customInstructions,
 					replaceInstructions: hookResult?.replaceInstructions ?? options?.replaceInstructions,

packages/agent/src/harness/compaction/branch-summarization.ts

@@ -1,5 +1,5 @@
-import type { Model } from "@earendil-works/pi-ai";
-import { completeSimple } from "@earendil-works/pi-ai";
+import type { Model, Models } from "@earendil-works/pi-ai";
+
 import type { AgentMessage } from "../../types.ts";
 import {
 	convertToLlm,
@@ -49,12 +49,10 @@ export interface CollectEntriesResult {
 
 /** Options for generating a branch summary. */
 export interface GenerateBranchSummaryOptions {
+	/** Provider collection the summarization request goes through; owns auth resolution. */
+	models: Models;
 	/** Model used for summarization. */
 	model: Model<any>;
-	/** API key forwarded to the provider. */
-	apiKey: string;
-	/** Optional request headers forwarded to the provider. */
-	headers?: Record<string, string>;
 	/** Abort signal for the summarization request. */
 	signal: AbortSignal;
 	/** Optional instructions appended to or replacing the default prompt. */
@@ -202,7 +200,7 @@ export async function generateBranchSummary(
 	entries: SessionTreeEntry[],
 	options: GenerateBranchSummaryOptions,
 ): Promise<Result<BranchSummaryResult, BranchSummaryError>> {
-	const { model, apiKey, headers, signal, customInstructions, replaceInstructions, reserveTokens = 16384 } = options;
+	const { models, model, signal, customInstructions, replaceInstructions, reserveTokens = 16384 } = options;
 	const contextWindow = model.contextWindow || 128000;
 	const tokenBudget = contextWindow - reserveTokens;
 
@@ -230,10 +228,10 @@ export async function generateBranchSummary(
 			timestamp: Date.now(),
 		},
 	];
-	const response = await completeSimple(
+	const response = await models.completeSimple(
 		model,
 		{ systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages },
-		{ apiKey, headers, signal, maxTokens: 2048 },
+		{ signal, maxTokens: 2048 },
 	);
 	if (response.stopReason === "aborted") {
 		return err(new BranchSummaryError("aborted", response.errorMessage || "Branch summary aborted"));

packages/agent/src/harness/compaction/compaction.ts

@@ -1,5 +1,4 @@
-import type { AssistantMessage, ImageContent, Model, TextContent, Usage } from "@earendil-works/pi-ai";
-import { completeSimple } from "@earendil-works/pi-ai";
+import type { AssistantMessage, ImageContent, Model, Models, TextContent, Usage } from "@earendil-works/pi-ai";
 import type { AgentMessage, ThinkingLevel } from "../../types.ts";
 import {
 	convertToLlm,
@@ -455,10 +454,9 @@ Keep each section concise. Preserve exact file paths, function names, and error
 /** Generate or update a conversation summary for compaction. */
 export async function generateSummary(
 	currentMessages: AgentMessage[],
+	models: Models,
 	model: Model<any>,
 	reserveTokens: number,
-	apiKey: string,
-	headers?: Record<string, string>,
 	signal?: AbortSignal,
 	customInstructions?: string,
 	previousSummary?: string,
@@ -490,10 +488,10 @@ export async function generateSummary(
 
 	const completionOptions =
 		model.reasoning && thinkingLevel && thinkingLevel !== "off"
-			? { maxTokens, signal, apiKey, headers, reasoning: thinkingLevel }
-			: { maxTokens, signal, apiKey, headers };
+			? { maxTokens, signal, reasoning: thinkingLevel }
+			: { maxTokens, signal };
 
-	const response = await completeSimple(
+	const response = await models.completeSimple(
 		model,
 		{ systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages },
 		completionOptions,
@@ -626,9 +624,8 @@ export { serializeConversation } from "./utils.ts";
 /** Generate compaction summary data from prepared session history. */
 export async function compact(
 	preparation: CompactionPreparation,
+	models: Models,
 	model: Model<any>,
-	apiKey: string,
-	headers?: Record<string, string>,
 	customInstructions?: string,
 	signal?: AbortSignal,
 	thinkingLevel?: ThinkingLevel,
@@ -655,25 +652,16 @@ export async function compact(
 			messagesToSummarize.length > 0
 				? generateSummary(
 						messagesToSummarize,
+						models,
 						model,
 						settings.reserveTokens,
-						apiKey,
-						headers,
 						signal,
 						customInstructions,
 						previousSummary,
 						thinkingLevel,
 					)
 				: Promise.resolve(ok<string, CompactionError>("No prior history.")),
-			generateTurnPrefixSummary(
-				turnPrefixMessages,
-				model,
-				settings.reserveTokens,
-				apiKey,
-				headers,
-				signal,
-				thinkingLevel,
-			),
+			generateTurnPrefixSummary(turnPrefixMessages, models, model, settings.reserveTokens, signal, thinkingLevel),
 		]);
 		if (!historyResult.ok) return err(historyResult.error);
 		if (!turnPrefixResult.ok) return err(turnPrefixResult.error);
@@ -681,10 +669,9 @@ export async function compact(
 	} else {
 		const summaryResult = await generateSummary(
 			messagesToSummarize,
+			models,
 			model,
 			settings.reserveTokens,
-			apiKey,
-			headers,
 			signal,
 			customInstructions,
 			previousSummary,
@@ -706,10 +693,9 @@ export async function compact(
 }
 async function generateTurnPrefixSummary(
 	messages: AgentMessage[],
+	models: Models,
 	model: Model<any>,
 	reserveTokens: number,
-	apiKey: string,
-	headers?: Record<string, string>,
 	signal?: AbortSignal,
 	thinkingLevel?: ThinkingLevel,
 ): Promise<Result<string, CompactionError>> {
@@ -728,12 +714,12 @@ async function generateTurnPrefixSummary(
 		},
 	];
 
-	const response = await completeSimple(
+	const response = await models.completeSimple(
 		model,
 		{ systemPrompt: SUMMARIZATION_SYSTEM_PROMPT, messages: summarizationMessages },
 		model.reasoning && thinkingLevel && thinkingLevel !== "off"
-			? { maxTokens, signal, apiKey, headers, reasoning: thinkingLevel }
-			: { maxTokens, signal, apiKey, headers },
+			? { maxTokens, signal, reasoning: thinkingLevel }
+			: { maxTokens, signal },
 	);
 	if (response.stopReason === "aborted") {
 		return err(new CompactionError("aborted", response.errorMessage || "Turn prefix summarization aborted"));

packages/agent/src/harness/types.ts

@@ -1,4 +1,4 @@
-import type { ImageContent, Model, SimpleStreamOptions, TextContent, Transport } from "@earendil-works/pi-ai";
+import type { ImageContent, Model, Models, SimpleStreamOptions, TextContent, Transport } from "@earendil-works/pi-ai";
 import type { AgentEvent, AgentMessage, AgentTool, QueueMode, ThinkingLevel } from "../index.ts";
 import type { Session } from "./session/session.ts";
 
@@ -802,6 +802,12 @@ export interface AgentHarnessOptions<
 > {
 	env: ExecutionEnv;
 	session: Session;
+	/**
+	 * Provider collection used for all model requests (turn streaming,
+	 * compaction, branch summarization). Auth resolves through the providers'
+	 * auth.
+	 */
+	models: Models;
 	tools?: TTool[];
 	/**
 	 * Concrete resources available to explicit invocation methods and system-prompt callbacks.
@@ -818,9 +824,6 @@ export interface AgentHarnessOptions<
 				activeTools: TTool[];
 				resources: AgentHarnessResources<TSkill, TPromptTemplate>;
 		  }) => string | Promise<string>);
-	getApiKeyAndHeaders?: (
-		model: Model<any>,
-	) => Promise<{ apiKey: string; headers?: Record<string, string> } | undefined>;
 	/** Curated stream/provider request options. Snapshotted at turn start. */
 	streamOptions?: AgentHarnessStreamOptions;
 	model: Model<any>;

packages/agent/src/types.ts

@@ -1,11 +1,13 @@
 import type {
+	Api,
 	AssistantMessage,
 	AssistantMessageEvent,
+	AssistantMessageEventStream,
+	Context,
 	ImageContent,
 	Message,
 	Model,
 	SimpleStreamOptions,
-	streamSimple,
 	TextContent,
 	Tool,
 	ToolResultMessage,
@@ -13,7 +15,8 @@ import type {
 import type { Static, TSchema } from "typebox";
 
 /**
- * Stream function used by the agent loop.
+ * Stream function used by the agent loop. `Models.streamSimple` satisfies
+ * this shape.
  *
  * Contract:
  * - Must not throw or return a rejected promise for request/model/runtime failures.
@@ -22,8 +25,10 @@ import type { Static, TSchema } from "typebox";
  *   final AssistantMessage with stopReason "error" or "aborted" and errorMessage.
  */
 export type StreamFn = (
-	...args: Parameters<typeof streamSimple>
-) => ReturnType<typeof streamSimple> | Promise<ReturnType<typeof streamSimple>>;
+	model: Model<Api>,
+	context: Context,
+	options?: SimpleStreamOptions,
+) => AssistantMessageEventStream | Promise<AssistantMessageEventStream>;
 
 /**
  * Configuration for how tool calls from a single assistant message are executed.

packages/agent/test/agent.test.ts

@@ -1,4 +1,4 @@
-import { type AssistantMessage, type AssistantMessageEvent, EventStream, getModel } from "@earendil-works/pi-ai";
+import { type AssistantMessage, type AssistantMessageEvent, EventStream, getModel } from "@earendil-works/pi-ai/compat";
 import { Type } from "typebox";
 import { describe, expect, it } from "vitest";
 import { Agent, type AgentEvent, type AgentTool, type AgentToolUpdateCallback } from "../src/index.ts";

packages/agent/test/harness/agent-harness-stream.test.ts

@@ -1,18 +1,27 @@
-import { fauxAssistantMessage, fauxToolCall, registerFauxProvider, type StreamOptions } from "@earendil-works/pi-ai";
-import { afterEach, describe, expect, it } from "vitest";
+import {
+	createModels,
+	type FauxProviderHandle,
+	fauxAssistantMessage,
+	fauxProvider,
+	fauxToolCall,
+	type StreamOptions,
+} from "@earendil-works/pi-ai";
+import { describe, expect, it } from "vitest";
 import { AgentHarness } from "../../src/harness/agent-harness.ts";
 import { NodeExecutionEnv } from "../../src/harness/env/nodejs.ts";
 import { InMemorySessionStorage } from "../../src/harness/session/memory-storage.ts";
 import { Session } from "../../src/harness/session/session.ts";
 import { calculateTool } from "../utils/calculate.ts";
 
-const registrations: Array<{ unregister(): void }> = [];
+/** Shared collection; each faux provider gets a unique id so coexisting fakes route correctly. */
+const models = createModels();
+let fauxCount = 0;
 
-afterEach(() => {
-	for (const registration of registrations.splice(0)) {
-		registration.unregister();
-	}
-});
+function newFaux(): FauxProviderHandle {
+	const faux = fauxProvider({ provider: `faux-${++fauxCount}` });
+	models.setProvider(faux.provider);
+	return faux;
+}
 
 function createHarness(options: ConstructorParameters<typeof AgentHarness>[0]): AgentHarness {
 	return new AgentHarness(options);
@@ -27,10 +36,9 @@ function captureOptions(options: StreamOptions | undefined): StreamOptions {
 }
 
 describe("AgentHarness stream configuration", () => {
-	it("snapshots stream options and merges auth headers before provider request hooks", async () => {
+	it("snapshots stream options before provider request hooks", async () => {
 		let capturedOptions: StreamOptions | undefined;
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([
 			(_context, options) => {
 				capturedOptions = options;
@@ -40,6 +48,7 @@ describe("AgentHarness stream configuration", () => {
 
 		const session = new Session(new InMemorySessionStorage({ metadata: { id: "session-1", createdAt: "now" } }));
 		const harness = createHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session,
 			model: registration.getModel(),
@@ -51,12 +60,11 @@ describe("AgentHarness stream configuration", () => {
 				metadata: { base: true },
 				cacheRetention: "none",
 			},
-			getApiKeyAndHeaders: async () => ({ apiKey: "secret", headers: { "x-auth": "auth" } }),
 		});
 
 		harness.on("before_provider_request", (event) => {
 			expect(event.sessionId).toBe("session-1");
-			expect(event.streamOptions.headers).toEqual({ "x-base": "base", "x-auth": "auth" });
+			expect(event.streamOptions.headers).toEqual({ "x-base": "base" });
 			return {
 				streamOptions: {
 					headers: { "x-hook": "hook" },
@@ -68,21 +76,19 @@ describe("AgentHarness stream configuration", () => {
 		await harness.prompt("hello");
 
 		expect(capturedOptions).toMatchObject({
-			apiKey: "secret",
 			timeoutMs: 1000,
 			maxRetries: 2,
 			maxRetryDelayMs: 3000,
 			sessionId: "session-1",
 			cacheRetention: "none",
 		});
-		expect(capturedOptions?.headers).toEqual({ "x-base": "base", "x-auth": "auth", "x-hook": "hook" });
+		expect(capturedOptions?.headers).toEqual({ "x-base": "base", "x-hook": "hook" });
 		expect(capturedOptions?.metadata).toEqual({ base: true, hook: true });
 	});
 
 	it("chains provider request patches and supports deletion semantics", async () => {
 		let capturedOptions: StreamOptions | undefined;
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([
 			(_context, options) => {
 				capturedOptions = options;
@@ -91,6 +97,7 @@ describe("AgentHarness stream configuration", () => {
 		]);
 
 		const harness = createHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -133,8 +140,7 @@ describe("AgentHarness stream configuration", () => {
 
 	it("uses updated stream options for save-point snapshots without mutating the active request", async () => {
 		const capturedOptions: StreamOptions[] = [];
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([
 			(_context, options) => {
 				capturedOptions.push(captureOptions(options));
@@ -149,6 +155,7 @@ describe("AgentHarness stream configuration", () => {
 		]);
 
 		const harness = createHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -174,8 +181,7 @@ describe("AgentHarness stream configuration", () => {
 	it("chains provider payload hooks", async () => {
 		const seenPayloads: unknown[] = [];
 		let finalPayload: unknown;
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([
 			async (_context, options, _state, model) => {
 				finalPayload = await options?.onPayload?.({ steps: ["provider"] }, model);
@@ -184,6 +190,7 @@ describe("AgentHarness stream configuration", () => {
 		]);
 
 		const harness = createHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),

packages/agent/test/harness/agent-harness.test.ts

@@ -1,5 +1,13 @@
-import { fauxAssistantMessage, fauxToolCall, getModel, registerFauxProvider } from "@earendil-works/pi-ai";
-import { afterEach, describe, expect, it } from "vitest";
+import {
+	createModels,
+	type FauxProviderHandle,
+	fauxAssistantMessage,
+	fauxProvider,
+	fauxToolCall,
+	type RegisterFauxProviderOptions,
+} from "@earendil-works/pi-ai";
+import { getModel } from "@earendil-works/pi-ai/compat";
+import { describe, expect, it } from "vitest";
 import { AgentHarness } from "../../src/harness/agent-harness.ts";
 import { NodeExecutionEnv } from "../../src/harness/env/nodejs.ts";
 import { InMemorySessionStorage } from "../../src/harness/session/memory-storage.ts";
@@ -17,7 +25,15 @@ interface AppPromptTemplate extends PromptTemplate {
 	source: "project" | "user";
 }
 
-const registrations: Array<{ unregister(): void }> = [];
+/** Shared collection; each faux provider gets a unique id so coexisting fakes route correctly. */
+const models = createModels();
+let fauxCount = 0;
+
+function newFaux(options: RegisterFauxProviderOptions = {}): FauxProviderHandle {
+	const faux = fauxProvider({ provider: `faux-${++fauxCount}`, ...options });
+	models.setProvider(faux.provider);
+	return faux;
+}
 
 function textFromUserMessages(messages: Array<{ role: string; content: unknown }>): string[] {
 	return messages.flatMap((message) => {
@@ -44,18 +60,13 @@ function getReasoning(options: unknown): unknown {
 	return options.reasoning;
 }
 
-afterEach(() => {
-	for (const registration of registrations.splice(0)) {
-		registration.unregister();
-	}
-});
-
 describe("AgentHarness", () => {
 	it("constructs directly and exposes queue modes", () => {
 		const session = new Session(new InMemorySessionStorage());
 		const env = new NodeExecutionEnv({ cwd: process.cwd() });
 		const initialModel = getModel("anthropic", "claude-sonnet-4-5");
 		const harness = new AgentHarness({
+			models,
 			env,
 			session,
 			model: initialModel,
@@ -76,8 +87,7 @@ describe("AgentHarness", () => {
 	});
 
 	it("drains one queued steering message at a time and emits queue updates", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		const userCounts: number[] = [];
 		registration.setResponses([
 			(context) => {
@@ -94,6 +104,7 @@ describe("AgentHarness", () => {
 			},
 		]);
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -119,8 +130,7 @@ describe("AgentHarness", () => {
 	});
 
 	it("appends before_agent_start messages and persists them", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		let requestText: string[] = [];
 		registration.setResponses([
 			(context) => {
@@ -130,6 +140,7 @@ describe("AgentHarness", () => {
 		]);
 		const session = new Session(new InMemorySessionStorage());
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session,
 			model: registration.getModel(),
@@ -151,8 +162,7 @@ describe("AgentHarness", () => {
 	});
 
 	it("abort clears steer and follow-up queues but preserves next-turn messages", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		let releaseFirstResponse: (() => void) | undefined;
 		let abortedSignal: AbortSignal | undefined;
 		const firstResponseReleased = new Promise<void>((resolve) => {
@@ -171,6 +181,7 @@ describe("AgentHarness", () => {
 			},
 		]);
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -206,8 +217,7 @@ describe("AgentHarness", () => {
 	});
 
 	it("drains follow-up messages one at a time after the agent would otherwise stop", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		const userCounts: number[] = [];
 		registration.setResponses([
 			(context) => {
@@ -224,6 +234,7 @@ describe("AgentHarness", () => {
 			},
 		]);
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -249,11 +260,11 @@ describe("AgentHarness", () => {
 	});
 
 	it("settles thrown hook failures with persisted assistant error messages", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([() => fauxAssistantMessage("should not be used")]);
 		const session = new Session(new InMemorySessionStorage());
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session,
 			model: registration.getModel(),
@@ -280,13 +291,12 @@ describe("AgentHarness", () => {
 	});
 
 	it("refreshes model, thinking level, resources, system prompt, and active tools at save points", async () => {
-		const registration = registerFauxProvider({
+		const registration = newFaux({
 			models: [
 				{ id: "first", reasoning: true },
 				{ id: "second", reasoning: true },
 			],
 		});
-		registrations.push(registration);
 		const secondModel = registration.getModel("second");
 		if (!secondModel) throw new Error("missing second faux model");
 		const captured: Array<{ modelId: string; reasoning: unknown; systemPrompt: string; tools: string[] }> = [];
@@ -313,6 +323,7 @@ describe("AgentHarness", () => {
 			},
 		]);
 		const harness = new AgentHarness<Skill, PromptTemplate, AgentTool>({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -345,11 +356,11 @@ describe("AgentHarness", () => {
 	});
 
 	it("orders pending listener session writes after agent-emitted messages", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([() => fauxAssistantMessage("ok")]);
 		const session = new Session(new InMemorySessionStorage());
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session,
 			model: registration.getModel(),
@@ -376,11 +387,11 @@ describe("AgentHarness", () => {
 	});
 
 	it("waitForIdle waits for external run settlement and awaited listeners", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([() => fauxAssistantMessage("ok")]);
 		const barrier = deferred();
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session: new Session(new InMemorySessionStorage()),
 			model: registration.getModel(),
@@ -408,8 +419,7 @@ describe("AgentHarness", () => {
 	});
 
 	it("runs tool_call and tool_result hooks through the direct loop", async () => {
-		const registration = registerFauxProvider();
-		registrations.push(registration);
+		const registration = newFaux();
 		registration.setResponses([
 			() =>
 				fauxAssistantMessage(fauxToolCall("calculate", { expression: "2 + 2" }, { id: "call-1" }), {
@@ -418,6 +428,7 @@ describe("AgentHarness", () => {
 		]);
 		const session = new Session(new InMemorySessionStorage());
 		const harness = new AgentHarness({
+			models,
 			env: new NodeExecutionEnv({ cwd: process.cwd() }),
 			session,
 			model: registration.getModel(),
@@ -462,6 +473,7 @@ describe("AgentHarness", () => {
 		const inspectTool: AppTool = { ...calculateTool, name: "inspect", source: "builtin" };
 		const searchTool: AppTool = { ...calculateTool, name: "search", source: "extension" };
 		const harness = new AgentHarness<AppSkill, AppPromptTemplate, AppTool>({
+			models,
 			env,
 			session,
 			model,
@@ -530,11 +542,12 @@ describe("AgentHarness", () => {
 		const env = new NodeExecutionEnv({ cwd: process.cwd() });
 		const model = getModel("anthropic", "claude-sonnet-4-5");
 		expect(
-			() => new AgentHarness({ env, session, model, tools: [calculateTool], activeToolNames: ["missing"] }),
+			() => new AgentHarness({ env, session, models, model, tools: [calculateTool], activeToolNames: ["missing"] }),
 		).toThrow(/Unknown tool/);
 		expect(
 			() =>
 				new AgentHarness({
+					models,
 					env,
 					session,
 					model,
@@ -545,6 +558,7 @@ describe("AgentHarness", () => {
 		expect(
 			() =>
 				new AgentHarness({
+					models,
 					env,
 					session,
 					model,
@@ -558,7 +572,7 @@ describe("AgentHarness", () => {
 		const session = new Session(new InMemorySessionStorage());
 		const env = new NodeExecutionEnv({ cwd: process.cwd() });
 		const model = getModel("anthropic", "claude-sonnet-4-5");
-		const harness = new AgentHarness<AppSkill, AppPromptTemplate, AgentTool>({ env, session, model });
+		const harness = new AgentHarness<AppSkill, AppPromptTemplate, AgentTool>({ env, session, models, model });
 		const skill: AppSkill = {
 			name: "inspect",
 			description: "Inspect things",

packages/agent/test/harness/compaction.test.ts

@@ -1,13 +1,14 @@
 import {
 	type AssistantMessage,
-	type FauxProviderRegistration,
+	createModels,
+	type FauxProviderHandle,
 	fauxAssistantMessage,
+	fauxProvider,
 	type Message,
 	type Model,
-	registerFauxProvider,
 	type Usage,
 } from "@earendil-works/pi-ai";
-import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import { beforeEach, describe, expect, it } from "vitest";
 import {
 	type CompactionPreparation,
 	calculateContextTokens,
@@ -121,11 +122,13 @@ function createModelChangeEntry(provider: string, modelId: string, parentId: str
 	};
 }
 
-function createFauxModel(
-	reasoning: boolean,
-	maxTokens = 8192,
-): { faux: FauxProviderRegistration; model: Model<string> } {
-	const faux = registerFauxProvider({
+/** Shared collection; each faux provider gets a unique id so coexisting fakes route correctly. */
+const models = createModels();
+let fauxCount = 0;
+
+function createFauxModel(reasoning: boolean, maxTokens = 8192): { faux: FauxProviderHandle; model: Model<string> } {
+	const faux = fauxProvider({
+		provider: `faux-${++fauxCount}`,
 		models: [
 			{
 				id: reasoning ? "reasoning-model" : "non-reasoning-model",
@@ -135,18 +138,10 @@ function createFauxModel(
 			},
 		],
 	});
-	fauxRegistrations.push(faux);
+	models.setProvider(faux.provider);
 	return { faux, model: faux.getModel() };
 }
 
-const fauxRegistrations: FauxProviderRegistration[] = [];
-
-afterEach(() => {
-	while (fauxRegistrations.length > 0) {
-		fauxRegistrations.pop()?.unregister();
-	}
-});
-
 describe("harness compaction", () => {
 	beforeEach(() => {
 		nextId = 0;
@@ -445,19 +440,9 @@ describe("harness compaction", () => {
 			},
 		]);
 		getOrThrow(
-			await generateSummary(
-				messages,
-				reasoningModel,
-				2000,
-				"test-key",
-				undefined,
-				undefined,
-				undefined,
-				undefined,
-				"medium",
-			),
+			await generateSummary(messages, models, reasoningModel, 2000, undefined, undefined, undefined, "medium"),
 		);
-		expect(seenOptions[0]).toMatchObject({ reasoning: "medium", apiKey: "test-key" });
+		expect(seenOptions[0]).toMatchObject({ reasoning: "medium" });
 
 		const { faux: fauxOff, model: offModel } = createFauxModel(true);
 		fauxOff.setResponses([
@@ -466,9 +451,7 @@ describe("harness compaction", () => {
 				return fauxAssistantMessage("## Goal\nTest summary");
 			},
 		]);
-		getOrThrow(
-			await generateSummary(messages, offModel, 2000, "test-key", undefined, undefined, undefined, undefined, "off"),
-		);
+		getOrThrow(await generateSummary(messages, models, offModel, 2000, undefined, undefined, undefined, "off"));
 		expect(seenOptions[1]).not.toHaveProperty("reasoning");
 
 		const { faux: fauxNonReasoning, model: nonReasoningModel } = createFauxModel(false);
@@ -479,17 +462,7 @@ describe("harness compaction", () => {
 			},
 		]);
 		getOrThrow(
-			await generateSummary(
-				messages,
-				nonReasoningModel,
-				2000,
-				"test-key",
-				undefined,
-				undefined,
-				undefined,
-				undefined,
-				"medium",
-			),
+			await generateSummary(messages, models, nonReasoningModel, 2000, undefined, undefined, undefined, "medium"),
 		);
 		expect(seenOptions[2]).not.toHaveProperty("reasoning");
 	});
@@ -508,16 +481,7 @@ describe("harness compaction", () => {
 		]);
 
 		const summary = getOrThrow(
-			await generateSummary(
-				messages,
-				model,
-				2000,
-				"test-key",
-				{ "x-test": "yes" },
-				undefined,
-				"focus",
-				"old summary",
-			),
+			await generateSummary(messages, models, model, 2000, undefined, "focus", "old summary"),
 		);
 
 		expect(summary).toContain("Test summary");
@@ -529,7 +493,7 @@ describe("harness compaction", () => {
 		const messages: AgentMessage[] = [createUserMessage("Summarize this.")];
 		const { faux: errorFaux, model: errorModel } = createFauxModel(false);
 		errorFaux.setResponses([fauxAssistantMessage("", { stopReason: "error", errorMessage: "boom" })]);
-		const errorResult = await generateSummary(messages, errorModel, 2000, "test-key");
+		const errorResult = await generateSummary(messages, models, errorModel, 2000);
 		expect(errorResult).toMatchObject({
 			ok: false,
 			error: { code: "summarization_failed", message: "Summarization failed: boom" },
@@ -537,7 +501,7 @@ describe("harness compaction", () => {
 
 		const { faux: abortedFaux, model: abortedModel } = createFauxModel(false);
 		abortedFaux.setResponses([fauxAssistantMessage("", { stopReason: "aborted", errorMessage: "stopped" })]);
-		const abortedResult = await generateSummary(messages, abortedModel, 2000, "test-key");
+		const abortedResult = await generateSummary(messages, models, abortedModel, 2000);
 		expect(abortedResult).toMatchObject({ ok: false, error: { code: "aborted", message: "stopped" } });
 	});
 
@@ -565,7 +529,7 @@ describe("harness compaction", () => {
 			settings: { enabled: true, reserveTokens: 500000, keepRecentTokens: 20000 },
 		};
 
-		getOrThrow(await compact(preparation, model, "test-key"));
+		getOrThrow(await compact(preparation, models, model));
 
 		expect(seenOptions.map((options) => options?.maxTokens)).toEqual([128000, 128000]);
 	});
@@ -583,7 +547,7 @@ describe("harness compaction", () => {
 		};
 		const { faux: historyFaux, model: historyModel } = createFauxModel(false);
 		historyFaux.setResponses([fauxAssistantMessage("", { stopReason: "error", errorMessage: "history failed" })]);
-		expect(await compact(preparation, historyModel, "test-key")).toMatchObject({
+		expect(await compact(preparation, models, historyModel)).toMatchObject({
 			ok: false,
 			error: { code: "summarization_failed", message: "Summarization failed: history failed" },
 		});
@@ -591,8 +555,8 @@ describe("harness compaction", () => {
 		const { model: invalidModel } = createFauxModel(false);
 		const invalidResult = await compact(
 			{ ...preparation, messagesToSummarize: [], firstKeptEntryId: "" },
+			models,
 			invalidModel,
-			"test-key",
 		);
 		expect(invalidResult).toMatchObject({ ok: false, error: { code: "invalid_session" } });
 	});
@@ -617,7 +581,7 @@ describe("harness compaction", () => {
 			settings: { enabled: true, reserveTokens: 2000, keepRecentTokens: 20 },
 		};
 
-		getOrThrow(await compact(preparation, model, "test-key", undefined, undefined, undefined, "high"));
+		getOrThrow(await compact(preparation, models, model, undefined, undefined, "high"));
 
 		expect(seenOptions[0]).toMatchObject({ reasoning: "high" });
 	});
@@ -636,14 +600,14 @@ describe("harness compaction", () => {
 		const { faux, model } = createFauxModel(false);
 		faux.setResponses([fauxAssistantMessage("", { stopReason: "error", errorMessage: "prefix failed" })]);
 
-		expect(await compact(preparation, model, "test-key")).toMatchObject({
+		expect(await compact(preparation, models, model)).toMatchObject({
 			ok: false,
 			error: { code: "summarization_failed", message: "Turn prefix summarization failed: prefix failed" },
 		});
 
 		const { faux: abortedFaux, model: abortedModel } = createFauxModel(false);
 		abortedFaux.setResponses([fauxAssistantMessage("", { stopReason: "aborted", errorMessage: "prefix stopped" })]);
-		expect(await compact(preparation, abortedModel, "test-key")).toMatchObject({
+		expect(await compact(preparation, models, abortedModel)).toMatchObject({
 			ok: false,
 			error: { code: "aborted", message: "prefix stopped" },
 		});
@@ -662,7 +626,7 @@ describe("harness compaction", () => {
 		expect(preparation).toBeDefined();
 		const { faux, model } = createFauxModel(false);
 		faux.setResponses([fauxAssistantMessage("## Goal\nTest summary")]);
-		const result = getOrThrow(await compact(preparation!, model, "test-key"));
+		const result = getOrThrow(await compact(preparation!, models, model));
 		expect(result.summary.length).toBeGreaterThan(0);
 		expect(result.firstKeptEntryId).toBeTruthy();
 		expect(result.details).toBeDefined();

packages/agent/test/scratch/simple.ts

@@ -1,6 +1,8 @@
 import { homedir } from "node:os";
 import { join } from "node:path";
-import { getModel } from "@earendil-works/pi-ai";
+import { createModels } from "@earendil-works/pi-ai";
+import { getBuiltinModel } from "@earendil-works/pi-ai/providers/all";
+import { openaiProvider } from "@earendil-works/pi-ai/providers/openai";
 import { NodeExecutionEnv } from "../../src/harness/env/nodejs.ts";
 import { InMemorySessionStorage } from "../../src/harness/session/memory-storage.ts";
 import {
@@ -35,11 +37,15 @@ const { promptTemplates: sourcedPromptTemplates } = await loadSourcedPromptTempl
 	(promptTemplate, source) => ({ ...promptTemplate, source }),
 );
 
+const models = createModels();
+models.setProvider(openaiProvider());
+
 const session = new Session(new InMemorySessionStorage());
 const agent = new AgentHarness({
 	env,
 	session,
-	model: getModel("openai", "gpt-5.5"),
+	models,
+	model: getBuiltinModel("openai", "gpt-5.5"),
 	thinkingLevel: "low",
 	systemPrompt: ({ env, resources }) =>
 		[

packages/agent/vitest.config.ts

@@ -1,9 +1,19 @@
+import { fileURLToPath } from "node:url";
 import { defineConfig } from "vitest/config";
 
+const aiSrcIndex = fileURLToPath(new URL("../ai/src/index.ts", import.meta.url));
+const aiSrcCompat = fileURLToPath(new URL("../ai/src/compat.ts", import.meta.url));
+
 export default defineConfig({
 	test: {
 		globals: true,
 		environment: "node",
 		testTimeout: 30000, // 30 seconds for API calls
 	},
+	resolve: {
+		alias: [
+			{ find: /^@earendil-works\/pi-ai$/, replacement: aiSrcIndex },
+			{ find: /^@earendil-works\/pi-ai\/compat$/, replacement: aiSrcCompat },
+		],
+	},
 });

packages/agent/vitest.harness.config.ts

@@ -1,5 +1,9 @@
+import { fileURLToPath } from "node:url";
 import { defineConfig } from "vitest/config";
 
+const aiSrcIndex = fileURLToPath(new URL("../ai/src/index.ts", import.meta.url));
+const aiSrcCompat = fileURLToPath(new URL("../ai/src/compat.ts", import.meta.url));
+
 export default defineConfig({
 	test: {
 		globals: true,
@@ -15,4 +19,10 @@ export default defineConfig({
 			reportsDirectory: "coverage/harness",
 		},
 	},
+	resolve: {
+		alias: [
+			{ find: /^@earendil-works\/pi-ai$/, replacement: aiSrcIndex },
+			{ find: /^@earendil-works\/pi-ai\/compat$/, replacement: aiSrcCompat },
+		],
+	},
 });

packages/ai/CHANGELOG.md

@@ -2,9 +2,26 @@
 
 ## [Unreleased]
 
+### Breaking Changes
+
+- The root entrypoint (`@earendil-works/pi-ai`) is now core-only and side-effect free. The old global API moved to the temporary `@earendil-works/pi-ai/compat` entrypoint, a strict superset of the root: switching a file's import path is the only migration step. Moved symbols include `stream`/`complete`/`streamSimple`/`completeSimple`, `getModel`/`getModels`/`getProviders` (now deprecated aliases of `getBuiltinModel`/`getBuiltinModels`/`getBuiltinProviders` from `@earendil-works/pi-ai/providers/all`), `registerApiProvider`/`unregisterApiProviders`/`resetApiProviders`/`getApiProvider`, `getEnvApiKey`/`findEnvKeys`, `setBedrockProviderModule`, the per-API lazy stream wrappers (`anthropicMessagesApi`, ...), and the image-generation API.
+- Renamed the `Provider` type to `ProviderId`. `Provider` now names the runtime provider interface (id, name, auth, model listing, stream behavior).
+- API implementation modules moved from `src/providers/` to `@earendil-works/pi-ai/api/*`, renamed by API id (`anthropic` -> `api/anthropic-messages`, `google` -> `api/google-generative-ai`, `mistral` -> `api/mistral-conversations`, `amazon-bedrock` -> `api/bedrock-converse-stream`), each exporting exactly `stream` and `streamSimple`. The old per-impl export names (`streamAnthropic`, `streamSimpleAnthropic`, ...) are gone; the legacy package subpaths (`./anthropic`, `./google`, ...) keep working and point at the new modules.
+
+### Added
+
+- New `Models` runtime: `createModels()` builds an isolated provider collection with sync model reads (`getModels`/`getModel` return the last-known lists), an explicit async `refresh(provider?)` for dynamic providers, auth resolution (`getAuth`), and `stream`/`complete`/`streamSimple`/`completeSimple` that resolve auth through the owning provider. `createProvider()` builds providers from parts (single API implementation or a map dispatched on `model.api`; static `models` array plus an optional `refreshModels` fetcher with in-flight dedupe); `hasApi()` narrows dynamically listed models.
+- Provider auth substrate: `ProviderAuth` (`{ apiKey?, oauth? }`), one type-tagged credential per provider, `CredentialStore` (`read`/`modify`/`delete` with serialized writes; in-memory default), `envApiKeyAuth()`, `lazyOAuth()`, and injectable `AuthContext`. OAuth refresh runs under the store lock with double-checked expiry; a stored credential owns its provider (no silent env fallback).
+- One provider factory per built-in provider under `@earendil-works/pi-ai/providers/*` (e.g. `anthropicProvider()`, `openrouterProvider()`), plus `@earendil-works/pi-ai/providers/all` with `builtinProviders()`/`builtinModels()` and typed `getBuiltin*` catalog reads. Generated catalogs are split per provider, so importing one provider pulls one catalog; `sideEffects` metadata makes the package tree-shakeable.
+- OAuth flows (Anthropic, OpenAI Codex, GitHub Copilot) gained `OAuthAuth` adapters (`login`/`refresh`/`toAuth`) on unified `prompt()`/`notify()` login callbacks; Copilot's per-credential base URL is derived in `toAuth()`.
+- `fauxProvider()` returns a faux `Provider` for tests built on explicit `Models` collections.
+- Image generation mirrors the chat-side design: `createImagesModels()`/`ImagesProvider`/`createImagesProvider()` with sync model reads, explicit `refresh()`, provider-resolved auth, and never-rejecting `generateImages()`; `openrouterImagesProvider()` factory plus `builtinImagesProviders()`/`builtinImagesModels()` in `providers/all`. The `ImagesProvider` id type alias is renamed to `ImagesProviderId`; the old global image API stays on `/compat`.
+- When Amazon Bedrock rejects an unsupported data retention mode, the error now links the AWS data retention documentation ([#5561](https://github.com/earendil-works/pi/pull/5561) by [@unexge](https://github.com/unexge)).
+
 ### Fixed
 
 - Fixed OpenCode Go GLM-5.2 metadata to expose `xhigh` reasoning and send `reasoning_effort: "max"` ([#5967](https://github.com/earendil-works/pi/issues/5967)).
+- Fixed Claude Fable 5 thinking-off requests to omit Anthropic's unsupported `thinking.type: "disabled"` payload ([#5567](https://github.com/earendil-works/pi/pull/5567) by [@tmustier](https://github.com/tmustier)).
 
 ## [0.79.10] - 2026-06-22
 

packages/ai/package.json

@@ -5,42 +5,59 @@
 	"type": "module",
 	"main": "./dist/index.js",
 	"types": "./dist/index.d.ts",
+	"sideEffects": [
+		"./dist/compat.js",
+		"./dist/images.js",
+		"./dist/providers/images/register-builtins.js"
+	],
 	"exports": {
 		".": {
 			"types": "./dist/index.d.ts",
 			"import": "./dist/index.js"
 		},
+		"./compat": {
+			"types": "./dist/compat.d.ts",
+			"import": "./dist/compat.js"
+		},
+		"./providers/*": {
+			"types": "./dist/providers/*.d.ts",
+			"import": "./dist/providers/*.js"
+		},
+		"./api/*": {
+			"types": "./dist/api/*.d.ts",
+			"import": "./dist/api/*.js"
+		},
 		"./anthropic": {
-			"types": "./dist/providers/anthropic.d.ts",
-			"import": "./dist/providers/anthropic.js"
+			"types": "./dist/api/anthropic-messages.d.ts",
+			"import": "./dist/api/anthropic-messages.js"
 		},
 		"./azure-openai-responses": {
-			"types": "./dist/providers/azure-openai-responses.d.ts",
-			"import": "./dist/providers/azure-openai-responses.js"
+			"types": "./dist/api/azure-openai-responses.d.ts",
+			"import": "./dist/api/azure-openai-responses.js"
 		},
 		"./google": {
-			"types": "./dist/providers/google.d.ts",
-			"import": "./dist/providers/google.js"
+			"types": "./dist/api/google-generative-ai.d.ts",
+			"import": "./dist/api/google-generative-ai.js"
 		},
 		"./google-vertex": {
-			"types": "./dist/providers/google-vertex.d.ts",
-			"import": "./dist/providers/google-vertex.js"
+			"types": "./dist/api/google-vertex.d.ts",
+			"import": "./dist/api/google-vertex.js"
 		},
 		"./mistral": {
-			"types": "./dist/providers/mistral.d.ts",
-			"import": "./dist/providers/mistral.js"
+			"types": "./dist/api/mistral-conversations.d.ts",
+			"import": "./dist/api/mistral-conversations.js"
 		},
 		"./openai-codex-responses": {
-			"types": "./dist/providers/openai-codex-responses.d.ts",
-			"import": "./dist/providers/openai-codex-responses.js"
+			"types": "./dist/api/openai-codex-responses.d.ts",
+			"import": "./dist/api/openai-codex-responses.js"
 		},
 		"./openai-completions": {
-			"types": "./dist/providers/openai-completions.d.ts",
-			"import": "./dist/providers/openai-completions.js"
+			"types": "./dist/api/openai-completions.d.ts",
+			"import": "./dist/api/openai-completions.js"
 		},
 		"./openai-responses": {
-			"types": "./dist/providers/openai-responses.d.ts",
-			"import": "./dist/providers/openai-responses.js"
+			"types": "./dist/api/openai-responses.d.ts",
+			"import": "./dist/api/openai-responses.js"
 		},
 		"./oauth": {
 			"types": "./dist/oauth.d.ts",

packages/ai/scripts/generate-models.ts

@@ -1,6 +1,6 @@
 #!/usr/bin/env node
 
-import { writeFileSync } from "fs";
+import { readdirSync, rmSync, writeFileSync } from "fs";
 import { join, dirname } from "path";
 import { fileURLToPath } from "url";
 import {
@@ -8,7 +8,7 @@ import {
 	CLOUDFLARE_AI_GATEWAY_COMPAT_BASE_URL,
 	CLOUDFLARE_AI_GATEWAY_OPENAI_BASE_URL,
 	CLOUDFLARE_WORKERS_AI_BASE_URL,
-} from "../src/providers/cloudflare.ts";
+} from "../src/api/cloudflare.ts";
 import type { AnthropicMessagesCompat, Api, KnownProvider, Model, OpenAICompletionsCompat } from "../src/types.ts";
 
 const __filename = fileURLToPath(import.meta.url);
@@ -2111,62 +2111,80 @@ async function generateModels() {
 		}
 	}
 
-	// Generate TypeScript file
-	let output = `// This file is auto-generated by scripts/generate-models.ts
+	// Generate TypeScript files: one catalog per provider plus an aggregator
+	const generatedHeader = `// This file is auto-generated by scripts/generate-models.ts
 // Do not edit manually - run 'npm run generate-models' to update
 
-import type { Model } from "./types.ts";
-
-export const MODELS = {
 `;
+	const catalogConstName = (providerId: string) => `${providerId.toUpperCase().replace(/[^A-Z0-9]+/g, "_")}_MODELS`;
 
-	// Generate provider sections (sorted for deterministic output)
-	const sortedProviderIds = Object.keys(providers).sort();
-	for (const providerId of sortedProviderIds) {
-		const models = providers[providerId];
-		output += `\t${JSON.stringify(providerId)}: {\n`;
-
-		const sortedModelIds = Object.keys(models).sort();
-		for (const modelId of sortedModelIds) {
-			const model = models[modelId];
-			output += `\t\t"${model.id}": {\n`;
-			output += `\t\t\tid: "${model.id}",\n`;
-			output += `\t\t\tname: "${model.name}",\n`;
-			output += `\t\t\tapi: "${model.api}",\n`;
-			output += `\t\t\tprovider: "${model.provider}",\n`;
-			if (model.baseUrl !== undefined) {
-				output += `\t\t\tbaseUrl: "${model.baseUrl}",\n`;
-			}
-			if (model.headers) {
-				output += `\t\t\theaders: ${JSON.stringify(model.headers)},\n`;
-			}
-			if (model.compat) {
-				output += `			compat: ${JSON.stringify(model.compat)},
-`;
-			}
-			output += `\t\t\treasoning: ${model.reasoning},\n`;
-			if (model.thinkingLevelMap) {
-				output += `\t\t\tthinkingLevelMap: ${JSON.stringify(model.thinkingLevelMap)},\n`;
-			}
-			output += `\t\t\tinput: [${model.input.map(i => `"${i}"`).join(", ")}],\n`;
-			output += `\t\t\tcost: {\n`;
-			output += `\t\t\t\tinput: ${model.cost.input},\n`;
-			output += `\t\t\t\toutput: ${model.cost.output},\n`;
-			output += `\t\t\t\tcacheRead: ${model.cost.cacheRead},\n`;
-			output += `\t\t\t\tcacheWrite: ${model.cost.cacheWrite},\n`;
-			output += `\t\t\t},\n`;
-			output += `\t\t\tcontextWindow: ${model.contextWindow},\n`;
-			output += `\t\t\tmaxTokens: ${model.maxTokens},\n`;
-			output += `\t\t} satisfies Model<"${model.api}">,\n`;
+	function emitModel(model: Model<any>, indent: string): string {
+		let output = `${indent}"${model.id}": {\n`;
+		output += `${indent}\tid: "${model.id}",\n`;
+		output += `${indent}\tname: "${model.name}",\n`;
+		output += `${indent}\tapi: "${model.api}",\n`;
+		output += `${indent}\tprovider: "${model.provider}",\n`;
+		if (model.baseUrl !== undefined) {
+			output += `${indent}\tbaseUrl: "${model.baseUrl}",\n`;
 		}
-
-		output += `\t},\n`;
+		if (model.headers) {
+			output += `${indent}\theaders: ${JSON.stringify(model.headers)},\n`;
+		}
+		if (model.compat) {
+			output += `${indent}\tcompat: ${JSON.stringify(model.compat)},\n`;
+		}
+		output += `${indent}\treasoning: ${model.reasoning},\n`;
+		if (model.thinkingLevelMap) {
+			output += `${indent}\tthinkingLevelMap: ${JSON.stringify(model.thinkingLevelMap)},\n`;
+		}
+		output += `${indent}\tinput: [${model.input.map(i => `"${i}"`).join(", ")}],\n`;
+		output += `${indent}\tcost: {\n`;
+		output += `${indent}\t\tinput: ${model.cost.input},\n`;
+		output += `${indent}\t\toutput: ${model.cost.output},\n`;
+		output += `${indent}\t\tcacheRead: ${model.cost.cacheRead},\n`;
+		output += `${indent}\t\tcacheWrite: ${model.cost.cacheWrite},\n`;
+		output += `${indent}\t},\n`;
+		output += `${indent}\tcontextWindow: ${model.contextWindow},\n`;
+		output += `${indent}\tmaxTokens: ${model.maxTokens},\n`;
+		output += `${indent}} satisfies Model<"${model.api}">,\n`;
+		return output;
 	}
 
-	output += `} as const;
-`;
+	const sortedProviderIds = Object.keys(providers).sort();
+	const providersDir = join(packageRoot, "src/providers");
 
-	// Write file
+	// Remove stale per-provider catalogs
+	for (const entry of readdirSync(providersDir)) {
+		if (entry.endsWith(".models.ts")) {
+			rmSync(join(providersDir, entry));
+		}
+	}
+
+	// Per-provider catalogs (sorted for deterministic output)
+	for (const providerId of sortedProviderIds) {
+		const models = providers[providerId];
+		let output = generatedHeader;
+		output += `import type { Model } from "../types.ts";\n\n`;
+		output += `export const ${catalogConstName(providerId)} = {\n`;
+		const sortedModelIds = Object.keys(models).sort();
+		for (const modelId of sortedModelIds) {
+			output += emitModel(models[modelId], "\t");
+		}
+		output += `} as const;\n`;
+		writeFileSync(join(providersDir, `${providerId}.models.ts`), output);
+	}
+	console.log(`Generated ${sortedProviderIds.length} catalogs under src/providers/`);
+
+	// Aggregator
+	let output = generatedHeader;
+	for (const providerId of sortedProviderIds) {
+		output += `import { ${catalogConstName(providerId)} } from "./providers/${providerId}.models.ts";\n`;
+	}
+	output += `\nexport const MODELS = {\n`;
+	for (const providerId of sortedProviderIds) {
+		output += `\t${JSON.stringify(providerId)}: ${catalogConstName(providerId)},\n`;
+	}
+	output += `} as const;\n`;
 	writeFileSync(join(packageRoot, "src/models.generated.ts"), output);
 	console.log("Generated src/models.generated.ts");
 

packages/ai/src/api/anthropic-messages.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const anthropicMessagesApi = (): ProviderStreams => lazyApi(() => import("./anthropic-messages.ts"));

packages/ai/src/api/azure-openai-responses.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const azureOpenAIResponsesApi = (): ProviderStreams => lazyApi(() => import("./azure-openai-responses.ts"));

packages/ai/src/api/azure-openai-responses.ts

@@ -0,0 +1,299 @@
+import { AzureOpenAI } from "openai";
+import type { ResponseCreateParamsStreaming } from "openai/resources/responses/responses.js";
+import { clampThinkingLevel } from "../models.ts";
+import type {
+	Api,
+	AssistantMessage,
+	Context,
+	Model,
+	SimpleStreamOptions,
+	StreamFunction,
+	StreamOptions,
+} from "../types.ts";
+import { AssistantMessageEventStream } from "../utils/event-stream.ts";
+import { headersToRecord } from "../utils/headers.ts";
+import { getProviderEnvValue } from "../utils/provider-env.ts";
+import { clampOpenAIPromptCacheKey } from "./openai-prompt-cache.ts";
+import { convertResponsesMessages, convertResponsesTools, processResponsesStream } from "./openai-responses-shared.ts";
+import { buildBaseOptions } from "./simple-options.ts";
+
+const DEFAULT_AZURE_API_VERSION = "v1";
+const AZURE_TOOL_CALL_PROVIDERS = new Set(["openai", "openai-codex", "opencode", "azure-openai-responses"]);
+
+function parseDeploymentNameMap(value: string | undefined): Map<string, string> {
+	const map = new Map<string, string>();
+	if (!value) return map;
+	for (const entry of value.split(",")) {
+		const trimmed = entry.trim();
+		if (!trimmed) continue;
+		const [modelId, deploymentName] = trimmed.split("=", 2);
+		if (!modelId || !deploymentName) continue;
+		map.set(modelId.trim(), deploymentName.trim());
+	}
+	return map;
+}
+
+function resolveDeploymentName(model: Model<"azure-openai-responses">, options?: AzureOpenAIResponsesOptions): string {
+	if (options?.azureDeploymentName) {
+		return options.azureDeploymentName;
+	}
+	const mappedDeployment = parseDeploymentNameMap(
+		getProviderEnvValue("AZURE_OPENAI_DEPLOYMENT_NAME_MAP", options?.env),
+	).get(model.id);
+	return mappedDeployment || model.id;
+}
+
+function formatAzureOpenAIError(error: unknown): string {
+	if (error instanceof Error) {
+		const status = (error as Error & { status?: unknown }).status;
+		const statusCode = typeof status === "number" ? status : undefined;
+		if (statusCode !== undefined) {
+			return `Azure OpenAI API error (${statusCode}): ${error.message}`;
+		}
+		return error.message;
+	}
+	try {
+		return JSON.stringify(error);
+	} catch {
+		return String(error);
+	}
+}
+
+// Azure OpenAI Responses-specific options
+export interface AzureOpenAIResponsesOptions extends StreamOptions {
+	reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh";
+	reasoningSummary?: "auto" | "detailed" | "concise" | null;
+	azureApiVersion?: string;
+	azureResourceName?: string;
+	azureBaseUrl?: string;
+	azureDeploymentName?: string;
+}
+
+/**
+ * Generate function for Azure OpenAI Responses API
+ */
+export const stream: StreamFunction<"azure-openai-responses", AzureOpenAIResponsesOptions> = (
+	model: Model<"azure-openai-responses">,
+	context: Context,
+	options?: AzureOpenAIResponsesOptions,
+): AssistantMessageEventStream => {
+	const stream = new AssistantMessageEventStream();
+
+	// Start async processing
+	(async () => {
+		const deploymentName = resolveDeploymentName(model, options);
+
+		const output: AssistantMessage = {
+			role: "assistant",
+			content: [],
+			api: "azure-openai-responses" as Api,
+			provider: model.provider,
+			model: model.id,
+			usage: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+				totalTokens: 0,
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+			},
+			stopReason: "stop",
+			timestamp: Date.now(),
+		};
+
+		try {
+			// Create Azure OpenAI client
+			const apiKey = options?.apiKey;
+			if (!apiKey) {
+				throw new Error(`No API key for provider: ${model.provider}`);
+			}
+			const client = createClient(model, apiKey, options);
+			let params = buildParams(model, context, options, deploymentName);
+			const nextParams = await options?.onPayload?.(params, model);
+			if (nextParams !== undefined) {
+				params = nextParams as ResponseCreateParamsStreaming;
+			}
+			const requestOptions = {
+				...(options?.signal ? { signal: options.signal } : {}),
+				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),
+				maxRetries: options?.maxRetries ?? 0,
+			};
+			const { data: openaiStream, response } = await client.responses.create(params, requestOptions).withResponse();
+			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
+			stream.push({ type: "start", partial: output });
+
+			await processResponsesStream(openaiStream, output, stream, model);
+
+			if (options?.signal?.aborted) {
+				throw new Error("Request was aborted");
+			}
+
+			if (output.stopReason === "aborted" || output.stopReason === "error") {
+				throw new Error("An unknown error occurred");
+			}
+
+			stream.push({ type: "done", reason: output.stopReason, message: output });
+			stream.end();
+		} catch (error) {
+			for (const block of output.content) {
+				delete (block as { index?: number }).index;
+				// partialJson is only a streaming scratch buffer; never persist it.
+				delete (block as { partialJson?: string }).partialJson;
+			}
+			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
+			output.errorMessage = formatAzureOpenAIError(error);
+			stream.push({ type: "error", reason: output.stopReason, error: output });
+			stream.end();
+		}
+	})();
+
+	return stream;
+};
+
+export const streamSimple: StreamFunction<"azure-openai-responses", SimpleStreamOptions> = (
+	model: Model<"azure-openai-responses">,
+	context: Context,
+	options?: SimpleStreamOptions,
+): AssistantMessageEventStream => {
+	const apiKey = options?.apiKey;
+	if (!apiKey) {
+		throw new Error(`No API key for provider: ${model.provider}`);
+	}
+
+	const base = buildBaseOptions(model, options, apiKey);
+	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
+	const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
+
+	return stream(model, context, {
+		...base,
+		reasoningEffort,
+	} satisfies AzureOpenAIResponsesOptions);
+};
+
+function normalizeAzureBaseUrl(baseUrl: string): string {
+	const trimmed = baseUrl.trim().replace(/\/+$/, "");
+	let url: URL;
+	try {
+		url = new URL(trimmed);
+	} catch {
+		throw new Error(`Invalid Azure OpenAI base URL: ${baseUrl}`);
+	}
+
+	const isAzureHost =
+		url.hostname.endsWith(".openai.azure.com") || url.hostname.endsWith(".cognitiveservices.azure.com");
+	const normalizedPath = url.pathname.replace(/\/+$/, "");
+
+	// Ensure Azure hosts have /openai/v1 as base path so the AzureOpenAI SDK
+	// can append /deployments/<model>/... and ?api-version=v1 correctly.
+	if (isAzureHost && (normalizedPath === "" || normalizedPath === "/" || normalizedPath === "/openai")) {
+		url.pathname = "/openai/v1";
+		url.search = "";
+	}
+
+	return url.toString().replace(/\/+$/, "");
+}
+
+function buildDefaultBaseUrl(resourceName: string): string {
+	return `https://${resourceName}.openai.azure.com/openai/v1`;
+}
+
+function resolveAzureConfig(
+	model: Model<"azure-openai-responses">,
+	options?: AzureOpenAIResponsesOptions,
+): { baseUrl: string; apiVersion: string } {
+	const apiVersion =
+		options?.azureApiVersion ||
+		getProviderEnvValue("AZURE_OPENAI_API_VERSION", options?.env) ||
+		DEFAULT_AZURE_API_VERSION;
+
+	const baseUrl =
+		options?.azureBaseUrl?.trim() || getProviderEnvValue("AZURE_OPENAI_BASE_URL", options?.env)?.trim() || undefined;
+	const resourceName = options?.azureResourceName || getProviderEnvValue("AZURE_OPENAI_RESOURCE_NAME", options?.env);
+
+	let resolvedBaseUrl = baseUrl;
+
+	if (!resolvedBaseUrl && resourceName) {
+		resolvedBaseUrl = buildDefaultBaseUrl(resourceName);
+	}
+
+	if (!resolvedBaseUrl && model.baseUrl) {
+		resolvedBaseUrl = model.baseUrl;
+	}
+
+	if (!resolvedBaseUrl) {
+		throw new Error(
+			"Azure OpenAI base URL is required. Set AZURE_OPENAI_BASE_URL or AZURE_OPENAI_RESOURCE_NAME, or pass azureBaseUrl, azureResourceName, or model.baseUrl.",
+		);
+	}
+
+	return {
+		baseUrl: normalizeAzureBaseUrl(resolvedBaseUrl),
+		apiVersion,
+	};
+}
+
+function createClient(model: Model<"azure-openai-responses">, apiKey: string, options?: AzureOpenAIResponsesOptions) {
+	const headers = { ...model.headers };
+
+	if (options?.headers) {
+		Object.assign(headers, options.headers);
+	}
+
+	const { baseUrl, apiVersion } = resolveAzureConfig(model, options);
+
+	return new AzureOpenAI({
+		apiKey,
+		apiVersion,
+		dangerouslyAllowBrowser: true,
+		defaultHeaders: headers,
+		baseURL: baseUrl,
+	});
+}
+
+function buildParams(
+	model: Model<"azure-openai-responses">,
+	context: Context,
+	options: AzureOpenAIResponsesOptions | undefined,
+	deploymentName: string,
+) {
+	const messages = convertResponsesMessages(model, context, AZURE_TOOL_CALL_PROVIDERS);
+
+	const params: ResponseCreateParamsStreaming = {
+		model: deploymentName,
+		input: messages,
+		stream: true,
+		prompt_cache_key: clampOpenAIPromptCacheKey(options?.sessionId),
+		store: false,
+	};
+
+	if (options?.maxTokens) {
+		params.max_output_tokens = options?.maxTokens;
+	}
+
+	if (options?.temperature !== undefined) {
+		params.temperature = options?.temperature;
+	}
+
+	if (context.tools && context.tools.length > 0) {
+		params.tools = convertResponsesTools(context.tools);
+	}
+
+	if (model.reasoning) {
+		if (options?.reasoningEffort || options?.reasoningSummary) {
+			const effort = options?.reasoningEffort
+				? (model.thinkingLevelMap?.[options.reasoningEffort] ?? options.reasoningEffort)
+				: "medium";
+			params.reasoning = {
+				effort: effort as NonNullable<typeof params.reasoning>["effort"],
+				summary: options?.reasoningSummary || "auto",
+			};
+			params.include = ["reasoning.encrypted_content"];
+		} else if (model.thinkingLevelMap?.off !== null) {
+			params.reasoning = {
+				effort: (model.thinkingLevelMap?.off ?? "none") as NonNullable<typeof params.reasoning>["effort"],
+			};
+		}
+	}
+
+	return params;
+}

packages/ai/src/api/bedrock-converse-stream.lazy.ts

@@ -0,0 +1,30 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+/**
+ * Loads the bedrock implementation through a variable specifier so bundlers
+ * (browser smoke, Bun compile) cannot follow the import into the Node-only
+ * AWS SDK. The `.ts`/`.js` rewrite keeps the trick working from both source
+ * and built output.
+ */
+const importNodeOnlyApi = (specifier: string): Promise<unknown> => {
+	const runtimeSpecifier = import.meta.url.endsWith(".js") ? specifier.replace(/\.ts$/, ".js") : specifier;
+	return import(runtimeSpecifier);
+};
+
+let bedrockModuleOverride: ProviderStreams | undefined;
+
+/**
+ * Overrides the dynamically imported bedrock implementation. Used by the Bun
+ * binary build, where the variable-specifier import cannot be bundled; the
+ * build registers a statically imported module instead.
+ */
+export function setBedrockProviderModule(module: ProviderStreams): void {
+	bedrockModuleOverride = module;
+}
+
+export const bedrockConverseStreamApi = (): ProviderStreams =>
+	lazyApi(
+		async () =>
+			bedrockModuleOverride ?? ((await importNodeOnlyApi("./bedrock-converse-stream.ts")) as ProviderStreams),
+	);

packages/ai/src/api/google-generative-ai.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const googleGenerativeAIApi = (): ProviderStreams => lazyApi(() => import("./google-generative-ai.ts"));

packages/ai/src/api/google-generative-ai.ts

@@ -0,0 +1,504 @@
+import {
+	type GenerateContentConfig,
+	type GenerateContentParameters,
+	GoogleGenAI,
+	type ThinkingConfig,
+} from "@google/genai";
+import { calculateCost, clampThinkingLevel } from "../models.ts";
+import type {
+	Api,
+	AssistantMessage,
+	Context,
+	Model,
+	SimpleStreamOptions,
+	StreamFunction,
+	StreamOptions,
+	TextContent,
+	ThinkingBudgets,
+	ThinkingContent,
+	ThinkingLevel,
+	ToolCall,
+} from "../types.ts";
+import { AssistantMessageEventStream } from "../utils/event-stream.ts";
+import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
+import type { GoogleThinkingLevel } from "./google-shared.ts";
+import {
+	convertMessages,
+	convertTools,
+	isThinkingPart,
+	mapStopReason,
+	mapToolChoice,
+	retainThoughtSignature,
+} from "./google-shared.ts";
+import { buildBaseOptions } from "./simple-options.ts";
+
+export interface GoogleOptions extends StreamOptions {
+	toolChoice?: "auto" | "none" | "any";
+	thinking?: {
+		enabled: boolean;
+		budgetTokens?: number; // -1 for dynamic, 0 to disable
+		level?: GoogleThinkingLevel;
+	};
+}
+
+// Counter for generating unique tool call IDs
+let toolCallCounter = 0;
+
+export const stream: StreamFunction<"google-generative-ai", GoogleOptions> = (
+	model: Model<"google-generative-ai">,
+	context: Context,
+	options?: GoogleOptions,
+): AssistantMessageEventStream => {
+	const stream = new AssistantMessageEventStream();
+
+	(async () => {
+		const output: AssistantMessage = {
+			role: "assistant",
+			content: [],
+			api: "google-generative-ai" as Api,
+			provider: model.provider,
+			model: model.id,
+			usage: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+				totalTokens: 0,
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+			},
+			stopReason: "stop",
+			timestamp: Date.now(),
+		};
+
+		try {
+			const apiKey = options?.apiKey;
+			if (!apiKey) {
+				throw new Error(`No API key for provider: ${model.provider}`);
+			}
+			const client = createClient(model, apiKey, options?.headers);
+			let params = buildParams(model, context, options);
+			const nextParams = await options?.onPayload?.(params, model);
+			if (nextParams !== undefined) {
+				params = nextParams as GenerateContentParameters;
+			}
+			const googleStream = await client.models.generateContentStream(params);
+
+			stream.push({ type: "start", partial: output });
+			let currentBlock: TextContent | ThinkingContent | null = null;
+			const blocks = output.content;
+			const blockIndex = () => blocks.length - 1;
+			for await (const chunk of googleStream) {
+				// @google/genai documents GenerateContentResponse.responseId as an output-only field
+				// used to identify each response. Keep the first non-empty one from the stream.
+				output.responseId ||= chunk.responseId;
+				const candidate = chunk.candidates?.[0];
+				if (candidate?.content?.parts) {
+					for (const part of candidate.content.parts) {
+						if (part.text !== undefined) {
+							const isThinking = isThinkingPart(part);
+							if (
+								!currentBlock ||
+								(isThinking && currentBlock.type !== "thinking") ||
+								(!isThinking && currentBlock.type !== "text")
+							) {
+								if (currentBlock) {
+									if (currentBlock.type === "text") {
+										stream.push({
+											type: "text_end",
+											contentIndex: blocks.length - 1,
+											content: currentBlock.text,
+											partial: output,
+										});
+									} else {
+										stream.push({
+											type: "thinking_end",
+											contentIndex: blockIndex(),
+											content: currentBlock.thinking,
+											partial: output,
+										});
+									}
+								}
+								if (isThinking) {
+									currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
+									output.content.push(currentBlock);
+									stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
+								} else {
+									currentBlock = { type: "text", text: "" };
+									output.content.push(currentBlock);
+									stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
+								}
+							}
+							if (currentBlock.type === "thinking") {
+								currentBlock.thinking += part.text;
+								currentBlock.thinkingSignature = retainThoughtSignature(
+									currentBlock.thinkingSignature,
+									part.thoughtSignature,
+								);
+								stream.push({
+									type: "thinking_delta",
+									contentIndex: blockIndex(),
+									delta: part.text,
+									partial: output,
+								});
+							} else {
+								currentBlock.text += part.text;
+								currentBlock.textSignature = retainThoughtSignature(
+									currentBlock.textSignature,
+									part.thoughtSignature,
+								);
+								stream.push({
+									type: "text_delta",
+									contentIndex: blockIndex(),
+									delta: part.text,
+									partial: output,
+								});
+							}
+						}
+
+						if (part.functionCall) {
+							if (currentBlock) {
+								if (currentBlock.type === "text") {
+									stream.push({
+										type: "text_end",
+										contentIndex: blockIndex(),
+										content: currentBlock.text,
+										partial: output,
+									});
+								} else {
+									stream.push({
+										type: "thinking_end",
+										contentIndex: blockIndex(),
+										content: currentBlock.thinking,
+										partial: output,
+									});
+								}
+								currentBlock = null;
+							}
+
+							// Generate unique ID if not provided or if it's a duplicate
+							const providedId = part.functionCall.id;
+							const needsNewId =
+								!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
+							const toolCallId = needsNewId
+								? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
+								: providedId;
+
+							const toolCall: ToolCall = {
+								type: "toolCall",
+								id: toolCallId,
+								name: part.functionCall.name || "",
+								arguments: (part.functionCall.args as Record<string, any>) ?? {},
+								...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
+							};
+
+							output.content.push(toolCall);
+							stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
+							stream.push({
+								type: "toolcall_delta",
+								contentIndex: blockIndex(),
+								delta: JSON.stringify(toolCall.arguments),
+								partial: output,
+							});
+							stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
+						}
+					}
+				}
+
+				if (candidate?.finishReason) {
+					output.stopReason = mapStopReason(candidate.finishReason);
+					if (output.content.some((b) => b.type === "toolCall")) {
+						output.stopReason = "toolUse";
+					}
+				}
+
+				if (chunk.usageMetadata) {
+					output.usage = {
+						input:
+							(chunk.usageMetadata.promptTokenCount || 0) - (chunk.usageMetadata.cachedContentTokenCount || 0),
+						output:
+							(chunk.usageMetadata.candidatesTokenCount || 0) + (chunk.usageMetadata.thoughtsTokenCount || 0),
+						cacheRead: chunk.usageMetadata.cachedContentTokenCount || 0,
+						cacheWrite: 0,
+						totalTokens: chunk.usageMetadata.totalTokenCount || 0,
+						cost: {
+							input: 0,
+							output: 0,
+							cacheRead: 0,
+							cacheWrite: 0,
+							total: 0,
+						},
+					};
+					calculateCost(model, output.usage);
+				}
+			}
+
+			if (currentBlock) {
+				if (currentBlock.type === "text") {
+					stream.push({
+						type: "text_end",
+						contentIndex: blockIndex(),
+						content: currentBlock.text,
+						partial: output,
+					});
+				} else {
+					stream.push({
+						type: "thinking_end",
+						contentIndex: blockIndex(),
+						content: currentBlock.thinking,
+						partial: output,
+					});
+				}
+			}
+
+			if (options?.signal?.aborted) {
+				throw new Error("Request was aborted");
+			}
+
+			if (output.stopReason === "aborted" || output.stopReason === "error") {
+				throw new Error("An unknown error occurred");
+			}
+
+			stream.push({ type: "done", reason: output.stopReason, message: output });
+			stream.end();
+		} catch (error) {
+			// Remove internal index property used during streaming
+			for (const block of output.content) {
+				if ("index" in block) {
+					delete (block as { index?: number }).index;
+				}
+			}
+			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
+			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
+			stream.push({ type: "error", reason: output.stopReason, error: output });
+			stream.end();
+		}
+	})();
+
+	return stream;
+};
+
+export const streamSimple: StreamFunction<"google-generative-ai", SimpleStreamOptions> = (
+	model: Model<"google-generative-ai">,
+	context: Context,
+	options?: SimpleStreamOptions,
+): AssistantMessageEventStream => {
+	const apiKey = options?.apiKey;
+	if (!apiKey) {
+		throw new Error(`No API key for provider: ${model.provider}`);
+	}
+
+	const base = buildBaseOptions(model, options, apiKey);
+	if (!options?.reasoning) {
+		return stream(model, context, { ...base, thinking: { enabled: false } } satisfies GoogleOptions);
+	}
+
+	const clampedReasoning = clampThinkingLevel(model, options.reasoning);
+	const effort = (clampedReasoning === "off" ? "high" : clampedReasoning) as ClampedThinkingLevel;
+	const googleModel = model as Model<"google-generative-ai">;
+
+	if (isGemini3ProModel(googleModel) || isGemini3FlashModel(googleModel) || isGemma4Model(googleModel)) {
+		return stream(model, context, {
+			...base,
+			thinking: {
+				enabled: true,
+				level: getThinkingLevel(effort, googleModel),
+			},
+		} satisfies GoogleOptions);
+	}
+
+	return stream(model, context, {
+		...base,
+		thinking: {
+			enabled: true,
+			budgetTokens: getGoogleBudget(googleModel, effort, options.thinkingBudgets),
+		},
+	} satisfies GoogleOptions);
+};
+
+function createClient(
+	model: Model<"google-generative-ai">,
+	apiKey?: string,
+	optionsHeaders?: Record<string, string>,
+): GoogleGenAI {
+	const httpOptions: { baseUrl?: string; apiVersion?: string; headers?: Record<string, string> } = {};
+	if (model.baseUrl) {
+		httpOptions.baseUrl = model.baseUrl;
+		httpOptions.apiVersion = ""; // baseUrl already includes version path, don't append
+	}
+	if (model.headers || optionsHeaders) {
+		httpOptions.headers = { ...model.headers, ...optionsHeaders };
+	}
+
+	return new GoogleGenAI({
+		apiKey,
+		httpOptions: Object.keys(httpOptions).length > 0 ? httpOptions : undefined,
+	});
+}
+
+function buildParams(
+	model: Model<"google-generative-ai">,
+	context: Context,
+	options: GoogleOptions = {},
+): GenerateContentParameters {
+	const contents = convertMessages(model, context);
+
+	const generationConfig: GenerateContentConfig = {};
+	if (options.temperature !== undefined) {
+		generationConfig.temperature = options.temperature;
+	}
+	if (options.maxTokens !== undefined) {
+		generationConfig.maxOutputTokens = options.maxTokens;
+	}
+
+	const config: GenerateContentConfig = {
+		...(Object.keys(generationConfig).length > 0 && generationConfig),
+		...(context.systemPrompt && { systemInstruction: sanitizeSurrogates(context.systemPrompt) }),
+		...(context.tools && context.tools.length > 0 && { tools: convertTools(context.tools) }),
+	};
+
+	if (context.tools && context.tools.length > 0 && options.toolChoice) {
+		config.toolConfig = {
+			functionCallingConfig: {
+				mode: mapToolChoice(options.toolChoice),
+			},
+		};
+	} else {
+		config.toolConfig = undefined;
+	}
+
+	if (options.thinking?.enabled && model.reasoning) {
+		const thinkingConfig: ThinkingConfig = { includeThoughts: true };
+		if (options.thinking.level !== undefined) {
+			// Cast to any since our GoogleThinkingLevel mirrors Google's ThinkingLevel enum values
+			thinkingConfig.thinkingLevel = options.thinking.level as any;
+		} else if (options.thinking.budgetTokens !== undefined) {
+			thinkingConfig.thinkingBudget = options.thinking.budgetTokens;
+		}
+		config.thinkingConfig = thinkingConfig;
+	} else if (model.reasoning && options.thinking && !options.thinking.enabled) {
+		config.thinkingConfig = getDisabledThinkingConfig(model);
+	}
+
+	if (options.signal) {
+		if (options.signal.aborted) {
+			throw new Error("Request aborted");
+		}
+		config.abortSignal = options.signal;
+	}
+
+	const params: GenerateContentParameters = {
+		model: model.id,
+		contents,
+		config,
+	};
+
+	return params;
+}
+
+type ClampedThinkingLevel = Exclude<ThinkingLevel, "xhigh">;
+
+function isGemma4Model(model: Model<"google-generative-ai">): boolean {
+	return /gemma-?4/.test(model.id.toLowerCase());
+}
+
+function isGemini3ProModel(model: Model<"google-generative-ai">): boolean {
+	return /gemini-3(?:\.\d+)?-pro/.test(model.id.toLowerCase());
+}
+
+function isGemini3FlashModel(model: Model<"google-generative-ai">): boolean {
+	const id = model.id.toLowerCase();
+	return /gemini-3(?:\.\d+)?-flash/.test(id) || id === "gemini-flash-latest" || id === "gemini-flash-lite-latest";
+}
+
+function getDisabledThinkingConfig(model: Model<"google-generative-ai">): ThinkingConfig {
+	// Google docs: Gemini 3.1 Pro cannot disable thinking, and Gemini 3 Flash / Flash-Lite
+	// do not support full thinking-off either. For Gemini 3 models, use the lowest supported
+	// thinkingLevel without includeThoughts so hidden thinking remains invisible to pi.
+	if (isGemini3ProModel(model)) {
+		return { thinkingLevel: "LOW" as any };
+	}
+	if (isGemini3FlashModel(model)) {
+		return { thinkingLevel: "MINIMAL" as any };
+	}
+	if (isGemma4Model(model)) {
+		return { thinkingLevel: "MINIMAL" as any };
+	}
+
+	// Gemini 2.x supports disabling via thinkingBudget = 0.
+	return { thinkingBudget: 0 };
+}
+
+function getThinkingLevel(effort: ClampedThinkingLevel, model: Model<"google-generative-ai">): GoogleThinkingLevel {
+	if (isGemini3ProModel(model)) {
+		switch (effort) {
+			case "minimal":
+			case "low":
+				return "LOW";
+			case "medium":
+			case "high":
+				return "HIGH";
+		}
+	}
+	if (isGemma4Model(model)) {
+		switch (effort) {
+			case "minimal":
+			case "low":
+				return "MINIMAL";
+			case "medium":
+			case "high":
+				return "HIGH";
+		}
+	}
+	switch (effort) {
+		case "minimal":
+			return "MINIMAL";
+		case "low":
+			return "LOW";
+		case "medium":
+			return "MEDIUM";
+		case "high":
+			return "HIGH";
+	}
+}
+
+function getGoogleBudget(
+	model: Model<"google-generative-ai">,
+	effort: ClampedThinkingLevel,
+	customBudgets?: ThinkingBudgets,
+): number {
+	if (customBudgets?.[effort] !== undefined) {
+		return customBudgets[effort]!;
+	}
+
+	if (model.id.includes("2.5-pro")) {
+		const budgets: Record<ClampedThinkingLevel, number> = {
+			minimal: 128,
+			low: 2048,
+			medium: 8192,
+			high: 32768,
+		};
+		return budgets[effort];
+	}
+
+	if (model.id.includes("2.5-flash-lite")) {
+		const budgets: Record<ClampedThinkingLevel, number> = {
+			minimal: 512,
+			low: 2048,
+			medium: 8192,
+			high: 24576,
+		};
+		return budgets[effort];
+	}
+
+	if (model.id.includes("2.5-flash")) {
+		const budgets: Record<ClampedThinkingLevel, number> = {
+			minimal: 128,
+			low: 2048,
+			medium: 8192,
+			high: 24576,
+		};
+		return budgets[effort];
+	}
+
+	return -1;
+}

packages/ai/src/api/google-vertex.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const googleVertexApi = (): ProviderStreams => lazyApi(() => import("./google-vertex.ts"));

packages/ai/src/api/google-vertex.ts

@@ -0,0 +1,582 @@
+import {
+	type GenerateContentConfig,
+	type GenerateContentParameters,
+	GoogleGenAI,
+	type HttpOptions,
+	ResourceScope,
+	type ThinkingConfig,
+	ThinkingLevel,
+} from "@google/genai";
+import { calculateCost, clampThinkingLevel } from "../models.ts";
+import type {
+	Api,
+	AssistantMessage,
+	Context,
+	Model,
+	ThinkingLevel as PiThinkingLevel,
+	ProviderEnv,
+	SimpleStreamOptions,
+	StreamFunction,
+	StreamOptions,
+	TextContent,
+	ThinkingBudgets,
+	ThinkingContent,
+	ToolCall,
+} from "../types.ts";
+import { AssistantMessageEventStream } from "../utils/event-stream.ts";
+import { getProviderEnvValue } from "../utils/provider-env.ts";
+import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
+import type { GoogleThinkingLevel } from "./google-shared.ts";
+import {
+	convertMessages,
+	convertTools,
+	isThinkingPart,
+	mapStopReason,
+	mapToolChoice,
+	retainThoughtSignature,
+} from "./google-shared.ts";
+import { buildBaseOptions } from "./simple-options.ts";
+
+export interface GoogleVertexOptions extends StreamOptions {
+	toolChoice?: "auto" | "none" | "any";
+	thinking?: {
+		enabled: boolean;
+		budgetTokens?: number; // -1 for dynamic, 0 to disable
+		level?: GoogleThinkingLevel;
+	};
+	project?: string;
+	location?: string;
+}
+
+const API_VERSION = "v1";
+const GCP_VERTEX_CREDENTIALS_MARKER = "gcp-vertex-credentials";
+
+const THINKING_LEVEL_MAP: Record<GoogleThinkingLevel, ThinkingLevel> = {
+	THINKING_LEVEL_UNSPECIFIED: ThinkingLevel.THINKING_LEVEL_UNSPECIFIED,
+	MINIMAL: ThinkingLevel.MINIMAL,
+	LOW: ThinkingLevel.LOW,
+	MEDIUM: ThinkingLevel.MEDIUM,
+	HIGH: ThinkingLevel.HIGH,
+};
+
+// Counter for generating unique tool call IDs
+let toolCallCounter = 0;
+
+export const stream: StreamFunction<"google-vertex", GoogleVertexOptions> = (
+	model: Model<"google-vertex">,
+	context: Context,
+	options?: GoogleVertexOptions,
+): AssistantMessageEventStream => {
+	const stream = new AssistantMessageEventStream();
+
+	(async () => {
+		const output: AssistantMessage = {
+			role: "assistant",
+			content: [],
+			api: "google-vertex" as Api,
+			provider: model.provider,
+			model: model.id,
+			usage: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+				totalTokens: 0,
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+			},
+			stopReason: "stop",
+			timestamp: Date.now(),
+		};
+
+		try {
+			const apiKey = resolveApiKey(options);
+			// Create the client using either a Vertex API key, if provided, or ADC with project and location
+			const client = apiKey
+				? createClientWithApiKey(model, apiKey, options?.headers)
+				: createClient(model, resolveProject(options), resolveLocation(options), options?.headers, options?.env);
+			let params = buildParams(model, context, options);
+			const nextParams = await options?.onPayload?.(params, model);
+			if (nextParams !== undefined) {
+				params = nextParams as GenerateContentParameters;
+			}
+			const googleStream = await client.models.generateContentStream(params);
+
+			stream.push({ type: "start", partial: output });
+			let currentBlock: TextContent | ThinkingContent | null = null;
+			const blocks = output.content;
+			const blockIndex = () => blocks.length - 1;
+			for await (const chunk of googleStream) {
+				// Vertex uses the same @google/genai GenerateContentResponse type as Gemini.
+				// responseId is documented there as an output-only identifier for each response.
+				output.responseId ||= chunk.responseId;
+				const candidate = chunk.candidates?.[0];
+				if (candidate?.content?.parts) {
+					for (const part of candidate.content.parts) {
+						if (part.text !== undefined) {
+							const isThinking = isThinkingPart(part);
+							if (
+								!currentBlock ||
+								(isThinking && currentBlock.type !== "thinking") ||
+								(!isThinking && currentBlock.type !== "text")
+							) {
+								if (currentBlock) {
+									if (currentBlock.type === "text") {
+										stream.push({
+											type: "text_end",
+											contentIndex: blocks.length - 1,
+											content: currentBlock.text,
+											partial: output,
+										});
+									} else {
+										stream.push({
+											type: "thinking_end",
+											contentIndex: blockIndex(),
+											content: currentBlock.thinking,
+											partial: output,
+										});
+									}
+								}
+								if (isThinking) {
+									currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
+									output.content.push(currentBlock);
+									stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
+								} else {
+									currentBlock = { type: "text", text: "" };
+									output.content.push(currentBlock);
+									stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
+								}
+							}
+							if (currentBlock.type === "thinking") {
+								currentBlock.thinking += part.text;
+								currentBlock.thinkingSignature = retainThoughtSignature(
+									currentBlock.thinkingSignature,
+									part.thoughtSignature,
+								);
+								stream.push({
+									type: "thinking_delta",
+									contentIndex: blockIndex(),
+									delta: part.text,
+									partial: output,
+								});
+							} else {
+								currentBlock.text += part.text;
+								currentBlock.textSignature = retainThoughtSignature(
+									currentBlock.textSignature,
+									part.thoughtSignature,
+								);
+								stream.push({
+									type: "text_delta",
+									contentIndex: blockIndex(),
+									delta: part.text,
+									partial: output,
+								});
+							}
+						}
+
+						if (part.functionCall) {
+							if (currentBlock) {
+								if (currentBlock.type === "text") {
+									stream.push({
+										type: "text_end",
+										contentIndex: blockIndex(),
+										content: currentBlock.text,
+										partial: output,
+									});
+								} else {
+									stream.push({
+										type: "thinking_end",
+										contentIndex: blockIndex(),
+										content: currentBlock.thinking,
+										partial: output,
+									});
+								}
+								currentBlock = null;
+							}
+
+							const providedId = part.functionCall.id;
+							const needsNewId =
+								!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
+							const toolCallId = needsNewId
+								? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
+								: providedId;
+
+							const toolCall: ToolCall = {
+								type: "toolCall",
+								id: toolCallId,
+								name: part.functionCall.name || "",
+								arguments: (part.functionCall.args as Record<string, any>) ?? {},
+								...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
+							};
+
+							output.content.push(toolCall);
+							stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
+							stream.push({
+								type: "toolcall_delta",
+								contentIndex: blockIndex(),
+								delta: JSON.stringify(toolCall.arguments),
+								partial: output,
+							});
+							stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
+						}
+					}
+				}
+
+				if (candidate?.finishReason) {
+					output.stopReason = mapStopReason(candidate.finishReason);
+					if (output.content.some((b) => b.type === "toolCall")) {
+						output.stopReason = "toolUse";
+					}
+				}
+
+				if (chunk.usageMetadata) {
+					output.usage = {
+						input:
+							(chunk.usageMetadata.promptTokenCount || 0) - (chunk.usageMetadata.cachedContentTokenCount || 0),
+						output:
+							(chunk.usageMetadata.candidatesTokenCount || 0) + (chunk.usageMetadata.thoughtsTokenCount || 0),
+						cacheRead: chunk.usageMetadata.cachedContentTokenCount || 0,
+						cacheWrite: 0,
+						totalTokens: chunk.usageMetadata.totalTokenCount || 0,
+						cost: {
+							input: 0,
+							output: 0,
+							cacheRead: 0,
+							cacheWrite: 0,
+							total: 0,
+						},
+					};
+					calculateCost(model, output.usage);
+				}
+			}
+
+			if (currentBlock) {
+				if (currentBlock.type === "text") {
+					stream.push({
+						type: "text_end",
+						contentIndex: blockIndex(),
+						content: currentBlock.text,
+						partial: output,
+					});
+				} else {
+					stream.push({
+						type: "thinking_end",
+						contentIndex: blockIndex(),
+						content: currentBlock.thinking,
+						partial: output,
+					});
+				}
+			}
+
+			if (options?.signal?.aborted) {
+				throw new Error("Request was aborted");
+			}
+
+			if (output.stopReason === "aborted" || output.stopReason === "error") {
+				throw new Error("An unknown error occurred");
+			}
+
+			stream.push({ type: "done", reason: output.stopReason, message: output });
+			stream.end();
+		} catch (error) {
+			// Remove internal index property used during streaming
+			for (const block of output.content) {
+				if ("index" in block) {
+					delete (block as { index?: number }).index;
+				}
+			}
+			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
+			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
+			stream.push({ type: "error", reason: output.stopReason, error: output });
+			stream.end();
+		}
+	})();
+
+	return stream;
+};
+
+export const streamSimple: StreamFunction<"google-vertex", SimpleStreamOptions> = (
+	model: Model<"google-vertex">,
+	context: Context,
+	options?: SimpleStreamOptions,
+): AssistantMessageEventStream => {
+	const base = buildBaseOptions(model, options, undefined);
+	if (!options?.reasoning) {
+		return stream(model, context, {
+			...base,
+			thinking: { enabled: false },
+		} satisfies GoogleVertexOptions);
+	}
+
+	const clampedReasoning = clampThinkingLevel(model, options.reasoning);
+	const effort = (clampedReasoning === "off" ? "high" : clampedReasoning) as ClampedThinkingLevel;
+	const geminiModel = model as unknown as Model<"google-generative-ai">;
+
+	if (isGemini3ProModel(geminiModel) || isGemini3FlashModel(geminiModel)) {
+		return stream(model, context, {
+			...base,
+			thinking: {
+				enabled: true,
+				level: getGemini3ThinkingLevel(effort, geminiModel),
+			},
+		} satisfies GoogleVertexOptions);
+	}
+
+	return stream(model, context, {
+		...base,
+		thinking: {
+			enabled: true,
+			budgetTokens: getGoogleBudget(geminiModel, effort, options.thinkingBudgets),
+		},
+	} satisfies GoogleVertexOptions);
+};
+
+function createClient(
+	model: Model<"google-vertex">,
+	project: string,
+	location: string,
+	optionsHeaders?: Record<string, string>,
+	env?: ProviderEnv,
+): GoogleGenAI {
+	const googleAuthOptions = buildGoogleAuthOptions(env);
+	return new GoogleGenAI({
+		vertexai: true,
+		project,
+		location,
+		apiVersion: API_VERSION,
+		...(googleAuthOptions ? { googleAuthOptions } : {}),
+		httpOptions: buildHttpOptions(model, optionsHeaders),
+	});
+}
+
+function createClientWithApiKey(
+	model: Model<"google-vertex">,
+	apiKey: string,
+	optionsHeaders?: Record<string, string>,
+): GoogleGenAI {
+	return new GoogleGenAI({
+		vertexai: true,
+		apiKey,
+		apiVersion: API_VERSION,
+		httpOptions: buildHttpOptions(model, optionsHeaders),
+	});
+}
+
+function buildHttpOptions(
+	model: Model<"google-vertex">,
+	optionsHeaders?: Record<string, string>,
+): HttpOptions | undefined {
+	const httpOptions: HttpOptions = {};
+	const baseUrl = resolveCustomBaseUrl(model.baseUrl);
+	if (baseUrl) {
+		httpOptions.baseUrl = baseUrl;
+		httpOptions.baseUrlResourceScope = ResourceScope.COLLECTION;
+		if (baseUrlIncludesApiVersion(baseUrl)) {
+			httpOptions.apiVersion = "";
+		}
+	}
+
+	if (model.headers || optionsHeaders) {
+		httpOptions.headers = { ...model.headers, ...optionsHeaders };
+	}
+
+	return Object.keys(httpOptions).length > 0 ? httpOptions : undefined;
+}
+
+function resolveCustomBaseUrl(baseUrl: string): string | undefined {
+	const trimmed = baseUrl.trim();
+	if (!trimmed || trimmed.includes("{location}")) {
+		return undefined;
+	}
+	return trimmed;
+}
+
+function baseUrlIncludesApiVersion(baseUrl: string): boolean {
+	try {
+		const url = new URL(baseUrl);
+		return url.pathname.split("/").some((part) => /^v\d+(?:beta\d*)?$/.test(part));
+	} catch {
+		return /(?:^|\/)v\d+(?:beta\d*)?(?:\/|$)/.test(baseUrl);
+	}
+}
+
+function buildGoogleAuthOptions(env?: ProviderEnv): { keyFilename: string } | undefined {
+	const keyFilename = getProviderEnvValue("GOOGLE_APPLICATION_CREDENTIALS", env);
+	return keyFilename ? { keyFilename } : undefined;
+}
+
+function resolveApiKey(options?: GoogleVertexOptions): string | undefined {
+	const apiKey = options?.apiKey?.trim();
+	if (!apiKey || apiKey === GCP_VERTEX_CREDENTIALS_MARKER || isPlaceholderApiKey(apiKey)) {
+		return undefined;
+	}
+	return apiKey;
+}
+
+function isPlaceholderApiKey(apiKey: string): boolean {
+	return /^<[^>]+>$/.test(apiKey);
+}
+
+function resolveProject(options?: GoogleVertexOptions): string {
+	const project =
+		options?.project ||
+		getProviderEnvValue("GOOGLE_CLOUD_PROJECT", options?.env) ||
+		getProviderEnvValue("GCLOUD_PROJECT", options?.env);
+	if (!project) {
+		throw new Error(
+			"Vertex AI requires a project ID. Set GOOGLE_CLOUD_PROJECT/GCLOUD_PROJECT or pass project in options.",
+		);
+	}
+	return project;
+}
+
+function resolveLocation(options?: GoogleVertexOptions): string {
+	const location = options?.location || getProviderEnvValue("GOOGLE_CLOUD_LOCATION", options?.env);
+	if (!location) {
+		throw new Error("Vertex AI requires a location. Set GOOGLE_CLOUD_LOCATION or pass location in options.");
+	}
+	return location;
+}
+
+function buildParams(
+	model: Model<"google-vertex">,
+	context: Context,
+	options: GoogleVertexOptions = {},
+): GenerateContentParameters {
+	const contents = convertMessages(model, context);
+
+	const generationConfig: GenerateContentConfig = {};
+	if (options.temperature !== undefined) {
+		generationConfig.temperature = options.temperature;
+	}
+	if (options.maxTokens !== undefined) {
+		generationConfig.maxOutputTokens = options.maxTokens;
+	}
+
+	const config: GenerateContentConfig = {
+		...(Object.keys(generationConfig).length > 0 && generationConfig),
+		...(context.systemPrompt && { systemInstruction: sanitizeSurrogates(context.systemPrompt) }),
+		...(context.tools && context.tools.length > 0 && { tools: convertTools(context.tools) }),
+	};
+
+	if (context.tools && context.tools.length > 0 && options.toolChoice) {
+		config.toolConfig = {
+			functionCallingConfig: {
+				mode: mapToolChoice(options.toolChoice),
+			},
+		};
+	} else {
+		config.toolConfig = undefined;
+	}
+
+	if (options.thinking?.enabled && model.reasoning) {
+		const thinkingConfig: ThinkingConfig = { includeThoughts: true };
+		if (options.thinking.level !== undefined) {
+			thinkingConfig.thinkingLevel = THINKING_LEVEL_MAP[options.thinking.level];
+		} else if (options.thinking.budgetTokens !== undefined) {
+			thinkingConfig.thinkingBudget = options.thinking.budgetTokens;
+		}
+		config.thinkingConfig = thinkingConfig;
+	} else if (model.reasoning && options.thinking && !options.thinking.enabled) {
+		config.thinkingConfig = getDisabledThinkingConfig(model);
+	}
+
+	if (options.signal) {
+		if (options.signal.aborted) {
+			throw new Error("Request aborted");
+		}
+		config.abortSignal = options.signal;
+	}
+
+	const params: GenerateContentParameters = {
+		model: model.id,
+		contents,
+		config,
+	};
+
+	return params;
+}
+
+type ClampedThinkingLevel = Exclude<PiThinkingLevel, "xhigh">;
+
+function isGemini3ProModel(model: Model<"google-generative-ai">): boolean {
+	return /gemini-3(?:\.\d+)?-pro/.test(model.id.toLowerCase());
+}
+
+function isGemini3FlashModel(model: Model<"google-generative-ai">): boolean {
+	const id = model.id.toLowerCase();
+	return /gemini-3(?:\.\d+)?-flash/.test(id) || id === "gemini-flash-latest" || id === "gemini-flash-lite-latest";
+}
+
+function getDisabledThinkingConfig(model: Model<"google-vertex">): ThinkingConfig {
+	// Google docs: Gemini 3.1 Pro cannot disable thinking, and Gemini 3 Flash / Flash-Lite
+	// do not support full thinking-off either. For Gemini 3 models, use the lowest supported
+	// thinkingLevel without includeThoughts so hidden thinking remains invisible to pi.
+	const geminiModel = model as unknown as Model<"google-generative-ai">;
+	if (isGemini3ProModel(geminiModel)) {
+		return { thinkingLevel: ThinkingLevel.LOW };
+	}
+	if (isGemini3FlashModel(geminiModel)) {
+		return { thinkingLevel: ThinkingLevel.MINIMAL };
+	}
+
+	// Gemini 2.x supports disabling via thinkingBudget = 0.
+	return { thinkingBudget: 0 };
+}
+
+function getGemini3ThinkingLevel(
+	effort: ClampedThinkingLevel,
+	model: Model<"google-generative-ai">,
+): GoogleThinkingLevel {
+	if (isGemini3ProModel(model)) {
+		switch (effort) {
+			case "minimal":
+			case "low":
+				return "LOW";
+			case "medium":
+			case "high":
+				return "HIGH";
+		}
+	}
+	switch (effort) {
+		case "minimal":
+			return "MINIMAL";
+		case "low":
+			return "LOW";
+		case "medium":
+			return "MEDIUM";
+		case "high":
+			return "HIGH";
+	}
+}
+
+function getGoogleBudget(
+	model: Model<"google-generative-ai">,
+	effort: ClampedThinkingLevel,
+	customBudgets?: ThinkingBudgets,
+): number {
+	if (customBudgets?.[effort] !== undefined) {
+		return customBudgets[effort]!;
+	}
+
+	if (model.id.includes("2.5-pro")) {
+		const budgets: Record<ClampedThinkingLevel, number> = {
+			minimal: 128,
+			low: 2048,
+			medium: 8192,
+			high: 32768,
+		};
+		return budgets[effort];
+	}
+
+	if (model.id.includes("2.5-flash")) {
+		const budgets: Record<ClampedThinkingLevel, number> = {
+			minimal: 128,
+			low: 2048,
+			medium: 8192,
+			high: 24576,
+		};
+		return budgets[effort];
+	}
+
+	return -1;
+}

packages/ai/src/api/lazy.ts

@@ -0,0 +1,70 @@
+import type { Api, AssistantMessage, AssistantMessageEvent, Model, ProviderStreams } from "../types.ts";
+import { AssistantMessageEventStream } from "../utils/event-stream.ts";
+
+function createSetupErrorMessage(model: Model<Api>, error: unknown): AssistantMessage {
+	return {
+		role: "assistant",
+		content: [],
+		api: model.api,
+		provider: model.provider,
+		model: model.id,
+		usage: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+			totalTokens: 0,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+		},
+		stopReason: "error",
+		errorMessage: error instanceof Error ? error.message : String(error),
+		timestamp: Date.now(),
+	};
+}
+
+function forwardStream(target: AssistantMessageEventStream, source: AsyncIterable<AssistantMessageEvent>): void {
+	(async () => {
+		for await (const event of source) {
+			target.push(event);
+		}
+		target.end();
+	})();
+}
+
+/**
+ * Returns a stream synchronously while running async setup (auth resolution,
+ * lazy module loading) behind it. Setup failures terminate the stream with an
+ * error event.
+ */
+export function lazyStream(
+	model: Model<Api>,
+	setup: () => Promise<AsyncIterable<AssistantMessageEvent>>,
+): AssistantMessageEventStream {
+	const outer = new AssistantMessageEventStream();
+
+	setup()
+		.then((inner) => {
+			forwardStream(outer, inner);
+		})
+		.catch((error) => {
+			const message = createSetupErrorMessage(model, error);
+			outer.push({ type: "error", reason: "error", error: message });
+			outer.end(message);
+		});
+
+	return outer;
+}
+
+/**
+ * Wraps a dynamically imported API implementation module as `ProviderStreams`.
+ * The module loads on first stream call; the host's import cache deduplicates
+ * loads. Load failures terminate the returned stream with an error event.
+ */
+export function lazyApi(load: () => Promise<ProviderStreams>): ProviderStreams {
+	return {
+		stream: (model, context, options) =>
+			lazyStream(model, async () => (await load()).stream(model, context, options)),
+		streamSimple: (model, context, options) =>
+			lazyStream(model, async () => (await load()).streamSimple(model, context, options)),
+	};
+}

packages/ai/src/api/mistral-conversations.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const mistralConversationsApi = (): ProviderStreams => lazyApi(() => import("./mistral-conversations.ts"));

packages/ai/src/api/mistral-conversations.ts

@@ -0,0 +1,664 @@
+import { Mistral } from "@mistralai/mistralai";
+import type {
+	ChatCompletionStreamRequest,
+	ChatCompletionStreamRequestMessage,
+	CompletionEvent,
+	ContentChunk,
+	FunctionTool,
+} from "@mistralai/mistralai/models/components";
+import { calculateCost, clampThinkingLevel } from "../models.ts";
+import type {
+	AssistantMessage,
+	Context,
+	Message,
+	Model,
+	SimpleStreamOptions,
+	StopReason,
+	StreamFunction,
+	StreamOptions,
+	TextContent,
+	ThinkingContent,
+	Tool,
+	ToolCall,
+} from "../types.ts";
+import { AssistantMessageEventStream } from "../utils/event-stream.ts";
+import { shortHash } from "../utils/hash.ts";
+import { parseStreamingJson } from "../utils/json-parse.ts";
+import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
+import { buildBaseOptions } from "./simple-options.ts";
+import { transformMessages } from "./transform-messages.ts";
+
+const MISTRAL_TOOL_CALL_ID_LENGTH = 9;
+const MAX_MISTRAL_ERROR_BODY_CHARS = 4000;
+
+/**
+ * Provider-specific options for the Mistral API.
+ */
+type MistralReasoningEffort = "none" | "high";
+
+export interface MistralOptions extends StreamOptions {
+	toolChoice?: "auto" | "none" | "any" | "required" | { type: "function"; function: { name: string } };
+	promptMode?: "reasoning";
+	reasoningEffort?: MistralReasoningEffort;
+}
+
+/**
+ * Stream responses from Mistral using `chat.stream`.
+ */
+export const stream: StreamFunction<"mistral-conversations", MistralOptions> = (
+	model: Model<"mistral-conversations">,
+	context: Context,
+	options?: MistralOptions,
+): AssistantMessageEventStream => {
+	const stream = new AssistantMessageEventStream();
+
+	(async () => {
+		const output = createOutput(model);
+
+		try {
+			const apiKey = options?.apiKey;
+			if (!apiKey) {
+				throw new Error(`No API key for provider: ${model.provider}`);
+			}
+
+			// Intentionally per-request: avoids shared SDK mutable state across concurrent consumers.
+			const mistral = new Mistral({
+				apiKey,
+				serverURL: model.baseUrl,
+			});
+
+			const normalizeMistralToolCallId = createMistralToolCallIdNormalizer();
+			const transformedMessages = transformMessages(context.messages, model, (id) => normalizeMistralToolCallId(id));
+
+			let payload = buildChatPayload(model, context, transformedMessages, options);
+			const nextPayload = await options?.onPayload?.(payload, model);
+			if (nextPayload !== undefined) {
+				payload = nextPayload as ChatCompletionStreamRequest;
+			}
+			const mistralStream = await mistral.chat.stream(payload, buildRequestOptions(model, options));
+			stream.push({ type: "start", partial: output });
+			await consumeChatStream(model, output, stream, mistralStream);
+
+			if (options?.signal?.aborted) {
+				throw new Error("Request was aborted");
+			}
+
+			if (output.stopReason === "aborted" || output.stopReason === "error") {
+				throw new Error("An unknown error occurred");
+			}
+
+			stream.push({ type: "done", reason: output.stopReason, message: output });
+			stream.end();
+		} catch (error) {
+			for (const block of output.content) {
+				// partialArgs is only a streaming scratch buffer; never persist it.
+				delete (block as { partialArgs?: string }).partialArgs;
+			}
+			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
+			output.errorMessage = formatMistralError(error);
+			stream.push({ type: "error", reason: output.stopReason, error: output });
+			stream.end();
+		}
+	})();
+
+	return stream;
+};
+
+/**
+ * Maps provider-agnostic `SimpleStreamOptions` to Mistral options.
+ */
+export const streamSimple: StreamFunction<"mistral-conversations", SimpleStreamOptions> = (
+	model: Model<"mistral-conversations">,
+	context: Context,
+	options?: SimpleStreamOptions,
+): AssistantMessageEventStream => {
+	const apiKey = options?.apiKey;
+	if (!apiKey) {
+		throw new Error(`No API key for provider: ${model.provider}`);
+	}
+
+	const base = buildBaseOptions(model, options, apiKey);
+	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
+	const reasoning = clampedReasoning === "off" ? undefined : clampedReasoning;
+	const shouldUseReasoning = model.reasoning && reasoning !== undefined;
+
+	return stream(model, context, {
+		...base,
+		promptMode: shouldUseReasoning && usesPromptModeReasoning(model) ? "reasoning" : undefined,
+		reasoningEffort:
+			shouldUseReasoning && usesReasoningEffort(model) ? mapReasoningEffort(model, reasoning) : undefined,
+	} satisfies MistralOptions);
+};
+
+function createOutput(model: Model<"mistral-conversations">): AssistantMessage {
+	return {
+		role: "assistant",
+		content: [],
+		api: model.api,
+		provider: model.provider,
+		model: model.id,
+		usage: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+			totalTokens: 0,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+		},
+		stopReason: "stop",
+		timestamp: Date.now(),
+	};
+}
+
+function createMistralToolCallIdNormalizer(): (id: string) => string {
+	const idMap = new Map<string, string>();
+	const reverseMap = new Map<string, string>();
+
+	return (id: string): string => {
+		const existing = idMap.get(id);
+		if (existing) return existing;
+
+		let attempt = 0;
+		while (true) {
+			const candidate = deriveMistralToolCallId(id, attempt);
+			const owner = reverseMap.get(candidate);
+			if (!owner || owner === id) {
+				idMap.set(id, candidate);
+				reverseMap.set(candidate, id);
+				return candidate;
+			}
+			attempt++;
+		}
+	};
+}
+
+function deriveMistralToolCallId(id: string, attempt: number): string {
+	const normalized = id.replace(/[^a-zA-Z0-9]/g, "");
+	if (attempt === 0 && normalized.length === MISTRAL_TOOL_CALL_ID_LENGTH) return normalized;
+	const seedBase = normalized || id;
+	const seed = attempt === 0 ? seedBase : `${seedBase}:${attempt}`;
+	return shortHash(seed)
+		.replace(/[^a-zA-Z0-9]/g, "")
+		.slice(0, MISTRAL_TOOL_CALL_ID_LENGTH);
+}
+
+function formatMistralError(error: unknown): string {
+	if (error instanceof Error) {
+		const sdkError = error as Error & { statusCode?: unknown; body?: unknown };
+		const statusCode = typeof sdkError.statusCode === "number" ? sdkError.statusCode : undefined;
+		const bodyText = typeof sdkError.body === "string" ? sdkError.body.trim() : undefined;
+		if (statusCode !== undefined && bodyText) {
+			return `Mistral API error (${statusCode}): ${truncateErrorText(bodyText, MAX_MISTRAL_ERROR_BODY_CHARS)}`;
+		}
+		if (statusCode !== undefined) return `Mistral API error (${statusCode}): ${error.message}`;
+		return error.message;
+	}
+	return safeJsonStringify(error);
+}
+
+function truncateErrorText(text: string, maxChars: number): string {
+	if (text.length <= maxChars) return text;
+	return `${text.slice(0, maxChars)}... [truncated ${text.length - maxChars} chars]`;
+}
+
+function safeJsonStringify(value: unknown): string {
+	try {
+		const serialized = JSON.stringify(value);
+		return serialized === undefined ? String(value) : serialized;
+	} catch {
+		return String(value);
+	}
+}
+
+function buildRequestOptions(model: Model<"mistral-conversations">, options?: MistralOptions) {
+	const requestOptions: {
+		signal?: AbortSignal;
+		retries: { strategy: "none" };
+		headers?: Record<string, string>;
+	} = {
+		retries: { strategy: "none" },
+	};
+	if (options?.signal) requestOptions.signal = options.signal;
+
+	const headers: Record<string, string> = {};
+	if (model.headers) Object.assign(headers, model.headers);
+	if (options?.headers) Object.assign(headers, options.headers);
+
+	// Mistral infrastructure uses `x-affinity` for KV-cache reuse (prefix caching).
+	// Respect explicit caller-provided header values.
+	if (shouldUsePromptCaching(options) && !headers["x-affinity"]) {
+		headers["x-affinity"] = options.sessionId;
+	}
+
+	if (Object.keys(headers).length > 0) {
+		requestOptions.headers = headers;
+	}
+
+	return requestOptions;
+}
+
+function buildChatPayload(
+	model: Model<"mistral-conversations">,
+	context: Context,
+	messages: Message[],
+	options?: MistralOptions,
+): ChatCompletionStreamRequest {
+	const payload: ChatCompletionStreamRequest = {
+		model: model.id,
+		stream: true,
+		messages: toChatMessages(messages, model.input.includes("image")),
+	};
+
+	if (context.tools?.length) payload.tools = toFunctionTools(context.tools);
+	if (options?.temperature !== undefined) payload.temperature = options.temperature;
+	if (options?.maxTokens !== undefined) payload.maxTokens = options.maxTokens;
+	if (options?.toolChoice) payload.toolChoice = mapToolChoice(options.toolChoice);
+	if (options?.promptMode) payload.promptMode = options.promptMode;
+	if (options?.reasoningEffort) payload.reasoningEffort = options.reasoningEffort;
+	if (shouldUsePromptCaching(options)) payload.promptCacheKey = options.sessionId;
+
+	if (context.systemPrompt) {
+		payload.messages.unshift({
+			role: "system",
+			content: sanitizeSurrogates(context.systemPrompt),
+		});
+	}
+
+	return payload;
+}
+
+function shouldUsePromptCaching(options?: MistralOptions): options is MistralOptions & { sessionId: string } {
+	return options?.cacheRetention !== "none" && !!options?.sessionId;
+}
+
+function getMistralCachedPromptTokens(usage: unknown, promptTokens: number): number {
+	const rawUsage = usage as {
+		promptTokensDetails?: { cachedTokens?: unknown } | null;
+		prompt_tokens_details?: { cached_tokens?: unknown } | null;
+		promptTokenDetails?: { cachedTokens?: unknown } | null;
+		prompt_token_details?: { cached_tokens?: unknown } | null;
+		numCachedTokens?: unknown;
+		num_cached_tokens?: unknown;
+	};
+	const rawCachedTokens =
+		rawUsage.promptTokensDetails?.cachedTokens ??
+		rawUsage.prompt_tokens_details?.cached_tokens ??
+		rawUsage.promptTokenDetails?.cachedTokens ??
+		rawUsage.prompt_token_details?.cached_tokens ??
+		rawUsage.numCachedTokens ??
+		rawUsage.num_cached_tokens ??
+		0;
+	const cachedTokens = typeof rawCachedTokens === "number" && Number.isFinite(rawCachedTokens) ? rawCachedTokens : 0;
+	return Math.min(promptTokens, Math.max(0, cachedTokens));
+}
+
+async function consumeChatStream(
+	model: Model<"mistral-conversations">,
+	output: AssistantMessage,
+	stream: AssistantMessageEventStream,
+	mistralStream: AsyncIterable<CompletionEvent>,
+): Promise<void> {
+	let currentBlock: TextContent | ThinkingContent | null = null;
+	const blocks = output.content;
+	const blockIndex = () => blocks.length - 1;
+	const toolBlocksByKey = new Map<string, number>();
+
+	const finishCurrentBlock = (block?: typeof currentBlock) => {
+		if (!block) return;
+		if (block.type === "text") {
+			stream.push({
+				type: "text_end",
+				contentIndex: blockIndex(),
+				content: block.text,
+				partial: output,
+			});
+			return;
+		}
+		if (block.type === "thinking") {
+			stream.push({
+				type: "thinking_end",
+				contentIndex: blockIndex(),
+				content: block.thinking,
+				partial: output,
+			});
+		}
+	};
+
+	for await (const event of mistralStream) {
+		const chunk = event.data;
+		// Mistral's streamed CompletionChunk carries an id field. Keep the first non-empty one,
+		// mirroring how OpenAI-style streaming exposes a stable response identifier per stream.
+		output.responseId ||= chunk.id;
+
+		if (chunk.usage) {
+			const promptTokens = chunk.usage.promptTokens || 0;
+			const cachedPromptTokens = getMistralCachedPromptTokens(chunk.usage, promptTokens);
+
+			output.usage.input = Math.max(0, promptTokens - cachedPromptTokens);
+			output.usage.output = chunk.usage.completionTokens || 0;
+			output.usage.cacheRead = cachedPromptTokens;
+			output.usage.cacheWrite = 0;
+			output.usage.totalTokens =
+				chunk.usage.totalTokens ||
+				output.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;
+			calculateCost(model, output.usage);
+		}
+
+		const choice = chunk.choices[0];
+		if (!choice) continue;
+
+		if (choice.finishReason) {
+			output.stopReason = mapChatStopReason(choice.finishReason);
+		}
+
+		const delta = choice.delta;
+		if (delta.content !== null && delta.content !== undefined) {
+			const contentItems = typeof delta.content === "string" ? [delta.content] : delta.content;
+			for (const item of contentItems) {
+				if (typeof item === "string") {
+					const textDelta = sanitizeSurrogates(item);
+					if (!currentBlock || currentBlock.type !== "text") {
+						finishCurrentBlock(currentBlock);
+						currentBlock = { type: "text", text: "" };
+						output.content.push(currentBlock);
+						stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
+					}
+					currentBlock.text += textDelta;
+					stream.push({
+						type: "text_delta",
+						contentIndex: blockIndex(),
+						delta: textDelta,
+						partial: output,
+					});
+					continue;
+				}
+
+				if (item.type === "thinking") {
+					const deltaText = item.thinking
+						.map((part) => ("text" in part ? part.text : ""))
+						.filter((text) => text.length > 0)
+						.join("");
+					const thinkingDelta = sanitizeSurrogates(deltaText);
+					if (!thinkingDelta) continue;
+					if (!currentBlock || currentBlock.type !== "thinking") {
+						finishCurrentBlock(currentBlock);
+						currentBlock = { type: "thinking", thinking: "" };
+						output.content.push(currentBlock);
+						stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
+					}
+					currentBlock.thinking += thinkingDelta;
+					stream.push({
+						type: "thinking_delta",
+						contentIndex: blockIndex(),
+						delta: thinkingDelta,
+						partial: output,
+					});
+					continue;
+				}
+
+				if (item.type === "text") {
+					const textDelta = sanitizeSurrogates(item.text);
+					if (!currentBlock || currentBlock.type !== "text") {
+						finishCurrentBlock(currentBlock);
+						currentBlock = { type: "text", text: "" };
+						output.content.push(currentBlock);
+						stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
+					}
+					currentBlock.text += textDelta;
+					stream.push({
+						type: "text_delta",
+						contentIndex: blockIndex(),
+						delta: textDelta,
+						partial: output,
+					});
+				}
+			}
+		}
+
+		const toolCalls = delta.toolCalls || [];
+		for (const toolCall of toolCalls) {
+			if (currentBlock) {
+				finishCurrentBlock(currentBlock);
+				currentBlock = null;
+			}
+			const callId =
+				toolCall.id && toolCall.id !== "null"
+					? toolCall.id
+					: deriveMistralToolCallId(`toolcall:${toolCall.index ?? 0}`, 0);
+			const key = `${callId}:${toolCall.index || 0}`;
+			const existingIndex = toolBlocksByKey.get(key);
+			let block: (ToolCall & { partialArgs?: string }) | undefined;
+
+			if (existingIndex !== undefined) {
+				const existing = output.content[existingIndex];
+				if (existing?.type === "toolCall") {
+					block = existing as ToolCall & { partialArgs?: string };
+				}
+			}
+
+			if (!block) {
+				block = {
+					type: "toolCall",
+					id: callId,
+					name: toolCall.function.name,
+					arguments: {},
+					partialArgs: "",
+				};
+				output.content.push(block);
+				toolBlocksByKey.set(key, output.content.length - 1);
+				stream.push({ type: "toolcall_start", contentIndex: output.content.length - 1, partial: output });
+			}
+
+			const argsDelta =
+				typeof toolCall.function.arguments === "string"
+					? toolCall.function.arguments
+					: JSON.stringify(toolCall.function.arguments || {});
+			block.partialArgs = (block.partialArgs || "") + argsDelta;
+			block.arguments = parseStreamingJson<Record<string, unknown>>(block.partialArgs);
+			stream.push({
+				type: "toolcall_delta",
+				contentIndex: toolBlocksByKey.get(key)!,
+				delta: argsDelta,
+				partial: output,
+			});
+		}
+	}
+
+	finishCurrentBlock(currentBlock);
+	for (const index of toolBlocksByKey.values()) {
+		const block = output.content[index];
+		if (block.type !== "toolCall") continue;
+		const toolBlock = block as ToolCall & { partialArgs?: string };
+		toolBlock.arguments = parseStreamingJson<Record<string, unknown>>(toolBlock.partialArgs);
+		// Finalize in-place and strip the scratch buffer so replay only
+		// carries parsed arguments.
+		delete toolBlock.partialArgs;
+		stream.push({
+			type: "toolcall_end",
+			contentIndex: index,
+			toolCall: toolBlock,
+			partial: output,
+		});
+	}
+}
+
+function toFunctionTools(tools: Tool[]): Array<FunctionTool & { type: "function" }> {
+	return tools.map((tool) => ({
+		type: "function",
+		function: {
+			name: tool.name,
+			description: tool.description,
+			parameters: stripSymbolKeys(tool.parameters) as Record<string, unknown>,
+			strict: false,
+		},
+	}));
+}
+
+function stripSymbolKeys(value: unknown): unknown {
+	if (Array.isArray(value)) {
+		return value.map((item) => stripSymbolKeys(item));
+	}
+
+	if (value && typeof value === "object") {
+		const result: Record<string, unknown> = {};
+		for (const [key, entry] of Object.entries(value)) {
+			result[key] = stripSymbolKeys(entry);
+		}
+		return result;
+	}
+
+	return value;
+}
+
+function toChatMessages(messages: Message[], supportsImages: boolean): ChatCompletionStreamRequestMessage[] {
+	const result: ChatCompletionStreamRequestMessage[] = [];
+
+	for (const msg of messages) {
+		if (msg.role === "user") {
+			if (typeof msg.content === "string") {
+				result.push({ role: "user", content: sanitizeSurrogates(msg.content) });
+				continue;
+			}
+			const hadImages = msg.content.some((item) => item.type === "image");
+			const content: ContentChunk[] = msg.content
+				.filter((item) => item.type === "text" || supportsImages)
+				.map((item) => {
+					if (item.type === "text") return { type: "text", text: sanitizeSurrogates(item.text) };
+					return { type: "image_url", imageUrl: `data:${item.mimeType};base64,${item.data}` };
+				});
+			if (content.length > 0) {
+				result.push({ role: "user", content });
+				continue;
+			}
+			if (hadImages && !supportsImages) {
+				result.push({ role: "user", content: "(image omitted: model does not support images)" });
+			}
+			continue;
+		}
+
+		if (msg.role === "assistant") {
+			const contentParts: ContentChunk[] = [];
+			const toolCalls: Array<{ id: string; type: "function"; function: { name: string; arguments: string } }> = [];
+
+			for (const block of msg.content) {
+				if (block.type === "text") {
+					if (block.text.trim().length > 0) {
+						contentParts.push({ type: "text", text: sanitizeSurrogates(block.text) });
+					}
+					continue;
+				}
+				if (block.type === "thinking") {
+					if (block.thinking.trim().length > 0) {
+						contentParts.push({
+							type: "thinking",
+							thinking: [{ type: "text", text: sanitizeSurrogates(block.thinking) }],
+						});
+					}
+					continue;
+				}
+				toolCalls.push({
+					id: block.id,
+					type: "function",
+					function: { name: block.name, arguments: JSON.stringify(block.arguments || {}) },
+				});
+			}
+
+			const assistantMessage: ChatCompletionStreamRequestMessage = { role: "assistant" };
+			if (contentParts.length > 0) assistantMessage.content = contentParts;
+			if (toolCalls.length > 0) assistantMessage.toolCalls = toolCalls;
+			if (contentParts.length > 0 || toolCalls.length > 0) result.push(assistantMessage);
+			continue;
+		}
+
+		const toolContent: ContentChunk[] = [];
+		const textResult = msg.content
+			.filter((part) => part.type === "text")
+			.map((part) => (part.type === "text" ? sanitizeSurrogates(part.text) : ""))
+			.join("\n");
+		const hasImages = msg.content.some((part) => part.type === "image");
+		const toolText = buildToolResultText(textResult, hasImages, supportsImages, msg.isError);
+		toolContent.push({ type: "text", text: toolText });
+		for (const part of msg.content) {
+			if (!supportsImages) continue;
+			if (part.type !== "image") continue;
+			toolContent.push({
+				type: "image_url",
+				imageUrl: `data:${part.mimeType};base64,${part.data}`,
+			});
+		}
+		result.push({
+			role: "tool",
+			toolCallId: msg.toolCallId,
+			name: msg.toolName,
+			content: toolContent,
+		});
+	}
+
+	return result;
+}
+
+function buildToolResultText(text: string, hasImages: boolean, supportsImages: boolean, isError: boolean): string {
+	const trimmed = text.trim();
+	const errorPrefix = isError ? "[tool error] " : "";
+
+	if (trimmed.length > 0) {
+		const imageSuffix = hasImages && !supportsImages ? "\n[tool image omitted: model does not support images]" : "";
+		return `${errorPrefix}${trimmed}${imageSuffix}`;
+	}
+
+	if (hasImages) {
+		if (supportsImages) {
+			return isError ? "[tool error] (see attached image)" : "(see attached image)";
+		}
+		return isError
+			? "[tool error] (image omitted: model does not support images)"
+			: "(image omitted: model does not support images)";
+	}
+
+	return isError ? "[tool error] (no tool output)" : "(no tool output)";
+}
+
+function usesReasoningEffort(model: Model<"mistral-conversations">): boolean {
+	return model.id === "mistral-small-2603" || model.id === "mistral-small-latest" || model.id === "mistral-medium-3.5";
+}
+
+function usesPromptModeReasoning(model: Model<"mistral-conversations">): boolean {
+	return model.reasoning && !usesReasoningEffort(model);
+}
+
+function mapReasoningEffort(
+	model: Model<"mistral-conversations">,
+	level: Exclude<SimpleStreamOptions["reasoning"], undefined>,
+): MistralReasoningEffort {
+	return (model.thinkingLevelMap?.[level] ?? "high") as MistralReasoningEffort;
+}
+
+function mapToolChoice(
+	choice: MistralOptions["toolChoice"],
+): "auto" | "none" | "any" | "required" | { type: "function"; function: { name: string } } | undefined {
+	if (!choice) return undefined;
+	if (choice === "auto" || choice === "none" || choice === "any" || choice === "required") {
+		return choice as any;
+	}
+	return {
+		type: "function",
+		function: { name: choice.function.name },
+	};
+}
+
+function mapChatStopReason(reason: string | null): StopReason {
+	if (reason === null) return "stop";
+	switch (reason) {
+		case "stop":
+			return "stop";
+		case "length":
+		case "model_length":
+			return "length";
+		case "tool_calls":
+			return "toolUse";
+		case "error":
+			return "error";
+		default:
+			return "stop";
+	}
+}

packages/ai/src/api/openai-codex-responses.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const openAICodexResponsesApi = (): ProviderStreams => lazyApi(() => import("./openai-codex-responses.ts"));

packages/ai/src/api/openai-codex-responses.ts

@@ -195,7 +195,7 @@ function createSSEHeaderTimeout(): { signal: AbortSignal; clear: () => void; err
 // Main Stream Function
 // ============================================================================
 
-export const streamOpenAICodexResponses: StreamFunction<"openai-codex-responses", OpenAICodexResponsesOptions> = (
+export const stream: StreamFunction<"openai-codex-responses", OpenAICodexResponsesOptions> = (
 	model: Model<"openai-codex-responses">,
 	context: Context,
 	options?: OpenAICodexResponsesOptions,
@@ -408,7 +408,7 @@ export const streamOpenAICodexResponses: StreamFunction<"openai-codex-responses"
 	return stream;
 };
 
-export const streamSimpleOpenAICodexResponses: StreamFunction<"openai-codex-responses", SimpleStreamOptions> = (
+export const streamSimple: StreamFunction<"openai-codex-responses", SimpleStreamOptions> = (
 	model: Model<"openai-codex-responses">,
 	context: Context,
 	options?: SimpleStreamOptions,
@@ -422,7 +422,7 @@ export const streamSimpleOpenAICodexResponses: StreamFunction<"openai-codex-resp
 	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
 	const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
 
-	return streamOpenAICodexResponses(model, context, {
+	return stream(model, context, {
 		...base,
 		reasoningEffort,
 	} satisfies OpenAICodexResponsesOptions);

packages/ai/src/api/openai-completions.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const openAICompletionsApi = (): ProviderStreams => lazyApi(() => import("./openai-completions.ts"));

packages/ai/src/api/openai-completions.ts

@@ -133,7 +133,7 @@ function resolveCacheRetention(cacheRetention?: CacheRetention, env?: ProviderEn
 	return "short";
 }
 
-export const streamOpenAICompletions: StreamFunction<"openai-completions", OpenAICompletionsOptions> = (
+export const stream: StreamFunction<"openai-completions", OpenAICompletionsOptions> = (
 	model: Model<"openai-completions">,
 	context: Context,
 	options?: OpenAICompletionsOptions,
@@ -463,7 +463,7 @@ export const streamOpenAICompletions: StreamFunction<"openai-completions", OpenA
 	return stream;
 };
 
-export const streamSimpleOpenAICompletions: StreamFunction<"openai-completions", SimpleStreamOptions> = (
+export const streamSimple: StreamFunction<"openai-completions", SimpleStreamOptions> = (
 	model: Model<"openai-completions">,
 	context: Context,
 	options?: SimpleStreamOptions,
@@ -478,7 +478,7 @@ export const streamSimpleOpenAICompletions: StreamFunction<"openai-completions",
 	const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
 	const toolChoice = (options as OpenAICompletionsOptions | undefined)?.toolChoice;
 
-	return streamOpenAICompletions(model, context, {
+	return stream(model, context, {
 		...base,
 		reasoningEffort,
 		toolChoice,

packages/ai/src/api/openai-responses.lazy.ts

@@ -0,0 +1,4 @@
+import type { ProviderStreams } from "../types.ts";
+import { lazyApi } from "./lazy.ts";
+
+export const openAIResponsesApi = (): ProviderStreams => lazyApi(() => import("./openai-responses.ts"));

packages/ai/src/api/openai-responses.ts

@@ -80,7 +80,7 @@ export interface OpenAIResponsesOptions extends StreamOptions {
 /**
  * Generate function for OpenAI Responses API
  */
-export const streamOpenAIResponses: StreamFunction<"openai-responses", OpenAIResponsesOptions> = (
+export const stream: StreamFunction<"openai-responses", OpenAIResponsesOptions> = (
 	model: Model<"openai-responses">,
 	context: Context,
 	options?: OpenAIResponsesOptions,
@@ -161,7 +161,7 @@ export const streamOpenAIResponses: StreamFunction<"openai-responses", OpenAIRes
 	return stream;
 };
 
-export const streamSimpleOpenAIResponses: StreamFunction<"openai-responses", SimpleStreamOptions> = (
+export const streamSimple: StreamFunction<"openai-responses", SimpleStreamOptions> = (
 	model: Model<"openai-responses">,
 	context: Context,
 	options?: SimpleStreamOptions,
@@ -175,7 +175,7 @@ export const streamSimpleOpenAIResponses: StreamFunction<"openai-responses", Sim
 	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
 	const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
 
-	return streamOpenAIResponses(model, context, {
+	return stream(model, context, {
 		...base,
 		reasoningEffort,
 	} satisfies OpenAIResponsesOptions);

packages/ai/src/api/openrouter-images.lazy.ts

@@ -0,0 +1,10 @@
+import type { ImagesModel, ProviderImages } from "../types.ts";
+
+export const openrouterImagesApi = (): ProviderImages => ({
+	generateImages: async (model, context, options) =>
+		(await import("./openrouter-images.ts")).generateImages(
+			model as ImagesModel<"openrouter-images">,
+			context,
+			options,
+		),
+});

packages/ai/src/api/openrouter-images.ts

@@ -14,9 +14,9 @@ import type {
 	ImagesModel,
 	ImagesOptions,
 	TextContent,
-} from "../../types.ts";
-import { headersToRecord } from "../../utils/headers.ts";
-import { sanitizeSurrogates } from "../../utils/sanitize-unicode.ts";
+} from "../types.ts";
+import { headersToRecord } from "../utils/headers.ts";
+import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
 
 interface OpenRouterGeneratedImage {
 	image_url?: string | { url?: string };
@@ -34,7 +34,7 @@ type OpenRouterImageGenerationResponse = ChatCompletion & {
 	choices: OpenRouterImageGenerationChoice[];
 };
 
-export const generateImagesOpenRouter: ImagesFunction<"openrouter-images", ImagesOptions> = async (
+export const generateImages: ImagesFunction<"openrouter-images", ImagesOptions> = async (
 	model: ImagesModel<"openrouter-images">,
 	context: ImagesContext,
 	options?: ImagesOptions,

packages/ai/src/auth/context.ts

@@ -0,0 +1,45 @@
+import type { AuthContext } from "./types.ts";
+
+interface NodeFsModule {
+	access(path: string): Promise<void>;
+}
+
+interface NodeOsModule {
+	homedir(): string;
+}
+
+// Variable specifier so browser bundlers do not try to resolve node builtins.
+const importNodeModule = (specifier: string): Promise<unknown> => import(specifier);
+
+function getProcessEnv(): Record<string, string | undefined> | undefined {
+	const proc = (globalThis as { process?: { env?: Record<string, string | undefined> } }).process;
+	return proc?.env;
+}
+
+/**
+ * Default auth context: env vars from `process.env` (undefined in browsers),
+ * file existence via node:fs (always false in browsers).
+ */
+export function defaultProviderAuthContext(): AuthContext {
+	return {
+		async env(name: string): Promise<string | undefined> {
+			const value = getProcessEnv()?.[name];
+			return typeof value === "string" && value.trim().length > 0 ? value : undefined;
+		},
+
+		async fileExists(path: string): Promise<boolean> {
+			try {
+				const fs = (await importNodeModule("node:fs/promises")) as NodeFsModule;
+				let resolved = path;
+				if (resolved.startsWith("~")) {
+					const os = (await importNodeModule("node:os")) as NodeOsModule;
+					resolved = os.homedir() + resolved.slice(1);
+				}
+				await fs.access(resolved);
+				return true;
+			} catch {
+				return false;
+			}
+		},
+	};
+}

packages/ai/src/auth/credential-store.ts

@@ -0,0 +1,47 @@
+import type { Credential, CredentialStore } from "./types.ts";
+
+/**
+ * Default in-memory credential store. Apps inject persistent stores.
+ * Keyed by `Provider.id`, one credential per provider; see `CredentialStore`.
+ * Writes are serialized per provider through a promise chain.
+ */
+export class InMemoryCredentialStore implements CredentialStore {
+	private credentials = new Map<string, Credential>();
+	private chains = new Map<string, Promise<unknown>>();
+
+	/** Serialize tasks per provider id. */
+	private enqueue<T>(providerId: string, task: () => Promise<T>): Promise<T> {
+		const previous = this.chains.get(providerId) ?? Promise.resolve();
+		const next = (async () => {
+			await previous.catch(() => {});
+			return task();
+		})();
+		this.chains.set(
+			providerId,
+			next.catch(() => {}),
+		);
+		return next;
+	}
+
+	async read(providerId: string): Promise<Credential | undefined> {
+		return this.credentials.get(providerId);
+	}
+
+	modify(
+		providerId: string,
+		fn: (current: Credential | undefined) => Promise<Credential | undefined>,
+	): Promise<Credential | undefined> {
+		return this.enqueue(providerId, async () => {
+			const current = this.credentials.get(providerId);
+			const next = await fn(current);
+			if (next !== undefined) this.credentials.set(providerId, next);
+			return next ?? current;
+		});
+	}
+
+	delete(providerId: string): Promise<void> {
+		return this.enqueue(providerId, async () => {
+			this.credentials.delete(providerId);
+		});
+	}
+}

packages/ai/src/auth/helpers.ts

@@ -0,0 +1,46 @@
+import type { ApiKeyAuth, OAuthAuth } from "./types.ts";
+
+/**
+ * Standard api-key auth: a stored credential key wins, otherwise the first
+ * set env var resolves. Includes a `login` that prompts for the key.
+ * Providers with non-standard resolution (metadata, ambient files, IAM)
+ * write their own `ApiKeyAuth`.
+ */
+export function envApiKeyAuth(name: string, envVars: readonly string[]): ApiKeyAuth {
+	return {
+		name,
+		login: async (callbacks) => {
+			const key = await callbacks.prompt({ type: "secret", message: `Enter ${name}` });
+			return { type: "api-key", key };
+		},
+		resolve: async ({ ctx, credential }) => {
+			if (credential?.key) return { auth: { apiKey: credential.key }, source: "stored credential" };
+			for (const envVar of envVars) {
+				const value = await ctx.env(envVar);
+				if (value) return { auth: { apiKey: value }, source: envVar };
+			}
+			return undefined;
+		},
+	};
+}
+
+/**
+ * Wraps a dynamically imported `OAuthAuth` so provider definitions can
+ * advertise OAuth without importing the implementation. The flow loads on
+ * first `login`/`refresh`/`toAuth` call; callers keep Node-only flow code out
+ * of bundles by loading through a bundler-opaque dynamic import (variable
+ * specifier, see the bedrock lazy wrapper).
+ */
+export function lazyOAuth(input: { name: string; load: () => Promise<OAuthAuth> }): OAuthAuth {
+	let promise: Promise<OAuthAuth> | undefined;
+	const loaded = () => {
+		promise ??= input.load();
+		return promise;
+	};
+	return {
+		name: input.name,
+		login: async (callbacks) => (await loaded()).login(callbacks),
+		refresh: async (credential) => (await loaded()).refresh(credential),
+		toAuth: async (credential) => (await loaded()).toAuth(credential),
+	};
+}

packages/ai/src/auth/resolve.ts

@@ -0,0 +1,117 @@
+import type { Api, ImagesApi, ImagesModel, Model } from "../types.ts";
+import type {
+	ApiKeyAuth,
+	ApiKeyCredential,
+	AuthContext,
+	AuthResult,
+	Credential,
+	CredentialStore,
+	OAuthAuth,
+	OAuthCredential,
+	ProviderAuth,
+} from "./types.ts";
+
+export type ModelsErrorCode = "model_source" | "model_validation" | "provider" | "stream" | "auth" | "oauth";
+
+export class ModelsError extends Error {
+	readonly code: ModelsErrorCode;
+
+	constructor(code: ModelsErrorCode, message: string, options?: { cause?: unknown }) {
+		super(message, options);
+		this.name = "ModelsError";
+		this.code = code;
+	}
+}
+
+/** Model shape auth resolution receives: chat or image-generation models. */
+export type AuthModel = Model<Api> | ImagesModel<ImagesApi>;
+
+/**
+ * Auth resolution shared by the `Models` and `ImagesModels` collections.
+ * A stored credential owns the provider: ambient/env is consulted only when
+ * nothing is stored. No silent env fallback after a failed refresh or for a
+ * credential type without a matching handler.
+ */
+export async function resolveProviderAuth(
+	provider: { id: string; auth: ProviderAuth },
+	model: AuthModel,
+	credentials: CredentialStore,
+	authContext: AuthContext,
+): Promise<AuthResult | undefined> {
+	const stored = await readCredential(credentials, provider.id);
+	if (stored) {
+		if (stored.type === "oauth" && provider.auth.oauth) {
+			return resolveStoredOAuth(credentials, provider.id, provider.auth.oauth, stored);
+		}
+		if (stored.type === "api-key" && provider.auth.apiKey) {
+			return resolveApiKey(authContext, provider.auth.apiKey, model, stored);
+		}
+		return undefined;
+	}
+
+	// Ambient (env vars, AWS profiles, ADC files).
+	return provider.auth.apiKey ? resolveApiKey(authContext, provider.auth.apiKey, model, undefined) : undefined;
+}
+
+/**
+ * OAuth resolution with double-checked locking (same pattern as today's
+ * AuthStorage): valid tokens cost zero locks; expired tokens lock, re-check
+ * expiry under the lock, refresh once globally, and persist the rotated
+ * credential before release.
+ */
+async function resolveStoredOAuth(
+	credentials: CredentialStore,
+	providerId: string,
+	oauth: OAuthAuth,
+	stored: OAuthCredential,
+): Promise<AuthResult | undefined> {
+	let credential = stored;
+
+	if (Date.now() >= credential.expires) {
+		// Optimistic check said expired; the authoritative check runs under the lock.
+		let post: Credential | undefined;
+		try {
+			post = await credentials.modify(providerId, async (current) => {
+				if (current?.type !== "oauth") return undefined; // logged out meanwhile
+				if (Date.now() < current.expires) return undefined; // another process/request refreshed
+				try {
+					return await oauth.refresh(current);
+				} catch (error) {
+					throw new ModelsError("oauth", `OAuth refresh failed for ${providerId}`, { cause: error });
+				}
+			});
+		} catch (error) {
+			if (error instanceof ModelsError) throw error;
+			throw new ModelsError("auth", `Credential store modify failed for ${providerId}`, { cause: error });
+		}
+		if (post?.type !== "oauth") return undefined; // logged out meanwhile
+		credential = post;
+	}
+
+	try {
+		return { auth: await oauth.toAuth(credential), source: "OAuth" };
+	} catch (error) {
+		throw new ModelsError("oauth", `OAuth auth derivation failed for ${providerId}`, { cause: error });
+	}
+}
+
+async function resolveApiKey(
+	authContext: AuthContext,
+	apiKey: ApiKeyAuth,
+	model: AuthModel,
+	credential: ApiKeyCredential | undefined,
+): Promise<AuthResult | undefined> {
+	try {
+		return await apiKey.resolve({ model, ctx: authContext, credential });
+	} catch (error) {
+		throw new ModelsError("auth", `API key auth failed for provider ${model.provider}`, { cause: error });
+	}
+}
+
+async function readCredential(credentials: CredentialStore, providerId: string): Promise<Credential | undefined> {
+	try {
+		return await credentials.read(providerId);
+	} catch (error) {
+		throw new ModelsError("auth", `Credential store read failed for ${providerId}`, { cause: error });
+	}
+}

packages/ai/src/auth/types.ts

@@ -0,0 +1,180 @@
+import type { Api, ImagesApi, ImagesModel, Model } from "../types.ts";
+import type { OAuthCredentials } from "../utils/oauth/types.ts";
+
+/**
+ * Request auth for a single model request. If a value cannot be expressed as
+ * `apiKey`, `headers`, or `baseUrl`, it is provider config, not auth.
+ */
+export interface ModelAuth {
+	apiKey?: string;
+	headers?: Record<string, string>;
+	baseUrl?: string;
+}
+
+/**
+ * Stored api-key credential. `metadata` holds non-key values such as
+ * Cloudflare account/gateway ids.
+ */
+export interface ApiKeyCredential {
+	type: "api-key";
+	key?: string;
+	metadata?: Record<string, string>;
+}
+
+/** Stored OAuth credential (`access`, `refresh`, `expires` from OAuthCredentials). */
+export interface OAuthCredential extends OAuthCredentials {
+	type: "oauth";
+}
+
+/** One type-tagged credential per provider — the shape of today's auth.json. */
+export type Credential = ApiKeyCredential | OAuthCredential;
+
+/**
+ * App-owned credential storage, keyed by `Provider.id`, one credential per
+ * provider. `modify` is the only write path, so every mutation is a
+ * serialized read-modify-write; `Models.getAuth()` runs OAuth refresh inside
+ * `modify` so concurrent requests cannot double-refresh a rotated token. The
+ * app persists a credential after login via
+ * `modify(provider.id, async () => credential)`. Login/logout orchestration
+ * is app-owned.
+ *
+ * Error semantics: `read` resolves `undefined` for missing entries. Methods
+ * reject only on storage failure; `Models` wraps such rejections in
+ * `ModelsError` with code "auth". Best-effort stores that serve an in-memory
+ * view and record persistence errors internally (like coding-agent's
+ * AuthStorage) are valid implementations.
+ */
+export interface CredentialStore {
+	/**
+	 * Read the stored credential, possibly expired. Display/status use;
+	 * resolved request auth comes from `Models.getAuth()`.
+	 */
+	read(providerId: string): Promise<Credential | undefined>;
+
+	/**
+	 * Serialized write — the only write path. `fn` sees the current credential
+	 * because correct writes (refresh, login-during-refresh) depend on it;
+	 * return the new credential, or undefined to leave the entry unchanged.
+	 * Mutual exclusion per provider id, cross-process too where the backing
+	 * store supports it (e.g. a file lock). Resolves with the post-write
+	 * credential. Rejections from `fn` propagate.
+	 */
+	modify(
+		providerId: string,
+		fn: (current: Credential | undefined) => Promise<Credential | undefined>,
+	): Promise<Credential | undefined>;
+
+	/** Remove a credential (logout). Implementations serialize this against `modify`. */
+	delete(providerId: string): Promise<void>;
+}
+
+/** Environment access for auth resolution. Injectable for tests and browsers. */
+export interface AuthContext {
+	env(name: string): Promise<string | undefined>;
+	/** Check whether a file exists. Supports a leading `~`. Always false in browsers. */
+	fileExists(path: string): Promise<boolean>;
+}
+
+/** Result of resolving auth for a model. */
+export interface AuthResult {
+	auth: ModelAuth;
+	/** Human-readable label for status UI: "ANTHROPIC_API_KEY", "OAuth", "~/.aws/credentials". */
+	source?: string;
+}
+
+/**
+ * Prompt shown to the user during login. `signal` lets the flow cancel a
+ * pending prompt when an out-of-band event resolves the step, e.g. a
+ * `manual_code` prompt raced against a callback server, aborted when the
+ * callback wins.
+ */
+export type AuthPrompt = { signal?: AbortSignal } & (
+	| { type: "text"; message: string; placeholder?: string }
+	| { type: "secret"; message: string; placeholder?: string }
+	| { type: "select"; message: string; options: readonly { id: string; label: string; description?: string }[] }
+	| { type: "manual_code"; message: string; placeholder?: string }
+);
+
+export type AuthEvent =
+	| { type: "auth_url"; url: string; instructions?: string }
+	| {
+			type: "device_code";
+			userCode: string;
+			verificationUri: string;
+			intervalSeconds?: number;
+			expiresInSeconds?: number;
+	  }
+	| { type: "progress"; message: string };
+
+/**
+ * Login interaction callbacks serving both api-key and OAuth flows.
+ *
+ * `prompt()` returns the entered/selected string (`select` returns the option
+ * id). Rejects on cancel/abort. `signal` aborts the whole login flow;
+ * per-prompt cancellation uses `AuthPrompt.signal`.
+ */
+export interface AuthLoginCallbacks {
+	signal?: AbortSignal;
+
+	prompt(prompt: AuthPrompt): Promise<string>;
+	notify(event: AuthEvent): void;
+}
+
+/**
+ * Api-key auth: stored key/metadata plus ambient sources (env vars, AWS
+ * profiles, ADC files). Ambient-only providers omit `login`.
+ */
+export interface ApiKeyAuth {
+	/** Display name, e.g. "Anthropic API key". */
+	name: string;
+
+	/** Interactive setup (prompt for key/metadata). Absent = ambient-only. */
+	login?(callbacks: AuthLoginCallbacks): Promise<ApiKeyCredential>;
+
+	/**
+	 * Resolve auth from the stored credential and/or ambient sources, merging
+	 * per field (`credential.key ?? env("...")`, `metadata.accountId ?? env("...")`).
+	 * undefined = not configured. Receives the chat or image-generation model
+	 * the request is for (both carry `provider` and `baseUrl`).
+	 */
+	resolve(input: {
+		model: Model<Api> | ImagesModel<ImagesApi>;
+		ctx: AuthContext;
+		credential?: ApiKeyCredential;
+	}): Promise<AuthResult | undefined>;
+}
+
+/**
+ * OAuth auth. The `refresh`/`toAuth` split lets `Models` own the locked
+ * refresh pattern: `refresh` produces a credential, `toAuth` derives request
+ * auth from whatever credential ends up stored.
+ */
+export interface OAuthAuth {
+	/** Display name, e.g. "Anthropic (Claude Pro/Max)". */
+	name: string;
+
+	login(callbacks: AuthLoginCallbacks): Promise<OAuthCredential>;
+
+	/**
+	 * Exchange the refresh token. Network call; throws on failure
+	 * (invalid_grant etc.). `Models` runs this under the store lock.
+	 */
+	refresh(credential: OAuthCredential): Promise<OAuthCredential>;
+
+	/**
+	 * Side-effect-free derivation of request auth from a valid credential.
+	 * Covers per-credential baseUrl (GitHub Copilot). Async so lazy wrappers
+	 * can load the implementation on first use.
+	 */
+	toAuth(credential: OAuthCredential): Promise<ModelAuth>;
+}
+
+/**
+ * Provider auth. At least one of `apiKey`/`oauth` must be present: even
+ * ambient-credential providers and keyless local servers provide `apiKey`
+ * auth whose `resolve()` reports whether the provider is configured.
+ */
+export interface ProviderAuth {
+	apiKey?: ApiKeyAuth;
+	oauth?: OAuthAuth;
+}

packages/ai/src/bedrock-provider.ts

@@ -1,6 +1,6 @@
-import { streamBedrock, streamSimpleBedrock } from "./providers/amazon-bedrock.ts";
+import { stream, streamSimple } from "./api/bedrock-converse-stream.ts";
 
 export const bedrockProviderModule = {
-	streamBedrock,
-	streamSimpleBedrock,
+	stream,
+	streamSimple,
 };

packages/ai/src/compat.ts

@@ -0,0 +1,150 @@
+/**
+ * Temporary compatibility entrypoint preserving the old global pi-ai API
+ * surface: api-dispatch `stream()`/`complete()` with env API key injection,
+ * the api-registry, generated catalog reads (`getModel`/`getModels`/
+ * `getProviders`), per-API lazy stream wrappers, and image generation.
+ *
+ * Existing apps switch imports from "@earendil-works/pi-ai" to
+ * "@earendil-works/pi-ai/compat" unchanged; new code uses `createModels()`
+ * and the provider factories. This module is deleted with the coding-agent
+ * ModelManager migration.
+ */
+
+export * from "./api/anthropic-messages.lazy.ts";
+export * from "./api/azure-openai-responses.lazy.ts";
+export * from "./api/bedrock-converse-stream.lazy.ts";
+export * from "./api/google-generative-ai.lazy.ts";
+export * from "./api/google-vertex.lazy.ts";
+export * from "./api/mistral-conversations.lazy.ts";
+export * from "./api/openai-codex-responses.lazy.ts";
+export * from "./api/openai-completions.lazy.ts";
+export * from "./api/openai-responses.lazy.ts";
+export * from "./api-registry.ts";
+export * from "./env-api-keys.ts";
+export * from "./image-models.ts";
+export * from "./images.ts";
+export * from "./images-api-registry.ts";
+export * from "./index.ts";
+export * from "./providers/images/register-builtins.ts";
+
+import { anthropicMessagesApi } from "./api/anthropic-messages.lazy.ts";
+import { azureOpenAIResponsesApi } from "./api/azure-openai-responses.lazy.ts";
+import { bedrockConverseStreamApi } from "./api/bedrock-converse-stream.lazy.ts";
+import { googleGenerativeAIApi } from "./api/google-generative-ai.lazy.ts";
+import { googleVertexApi } from "./api/google-vertex.lazy.ts";
+import { mistralConversationsApi } from "./api/mistral-conversations.lazy.ts";
+import { openAICodexResponsesApi } from "./api/openai-codex-responses.lazy.ts";
+import { openAICompletionsApi } from "./api/openai-completions.lazy.ts";
+import { openAIResponsesApi } from "./api/openai-responses.lazy.ts";
+import { clearApiProviders, getApiProvider, registerApiProvider } from "./api-registry.ts";
+import { getEnvApiKey } from "./env-api-keys.ts";
+import { getBuiltinModel, getBuiltinModels, getBuiltinProviders } from "./providers/all.ts";
+import type {
+	Api,
+	AssistantMessage,
+	AssistantMessageEventStream,
+	Context,
+	Model,
+	ProviderStreamOptions,
+	ProviderStreams,
+	SimpleStreamOptions,
+	StreamOptions,
+} from "./types.ts";
+
+/** @deprecated Static catalog read. Use `getBuiltinModel` from "@earendil-works/pi-ai/providers/all" or `Models.getModel()`. */
+export const getModel = getBuiltinModel;
+
+/** @deprecated Static catalog read. Use `getBuiltinModels` from "@earendil-works/pi-ai/providers/all" or `Models.getModels()`. */
+export const getModels = getBuiltinModels;
+
+/** @deprecated Static catalog read. Use `getBuiltinProviders` from "@earendil-works/pi-ai/providers/all" or `Models.getProviders()`. */
+export const getProviders = getBuiltinProviders;
+
+const BUILTIN_APIS: [Api, ProviderStreams][] = [
+	["anthropic-messages", anthropicMessagesApi()],
+	["openai-completions", openAICompletionsApi()],
+	["openai-responses", openAIResponsesApi()],
+	["openai-codex-responses", openAICodexResponsesApi()],
+	["azure-openai-responses", azureOpenAIResponsesApi()],
+	["google-generative-ai", googleGenerativeAIApi()],
+	["google-vertex", googleVertexApi()],
+	["mistral-conversations", mistralConversationsApi()],
+	["bedrock-converse-stream", bedrockConverseStreamApi()],
+];
+
+/**
+ * Registers the builtin API implementations into the api-registry without
+ * clobbering existing entries: compat may load after a test or extension has
+ * already registered an override for a builtin api id.
+ */
+export function registerBuiltInApiProviders(): void {
+	for (const [api, streams] of BUILTIN_APIS) {
+		if (getApiProvider(api)) continue;
+		registerApiProvider({ api, stream: streams.stream, streamSimple: streams.streamSimple });
+	}
+}
+
+export function resetApiProviders(): void {
+	clearApiProviders();
+	registerBuiltInApiProviders();
+}
+
+registerBuiltInApiProviders();
+
+function hasExplicitApiKey(apiKey: string | undefined): apiKey is string {
+	return typeof apiKey === "string" && apiKey.trim().length > 0;
+}
+
+function withEnvApiKey<TOptions extends StreamOptions>(
+	model: Model<Api>,
+	options: TOptions | undefined,
+): TOptions | undefined {
+	if (hasExplicitApiKey(options?.apiKey)) return options;
+	const apiKey = getEnvApiKey(model.provider);
+	if (!apiKey) return options;
+	return { ...options, apiKey } as TOptions;
+}
+
+function resolveApiProvider(api: Api) {
+	const provider = getApiProvider(api);
+	if (!provider) {
+		throw new Error(`No API provider registered for api: ${api}`);
+	}
+	return provider;
+}
+
+export function stream<TApi extends Api>(
+	model: Model<TApi>,
+	context: Context,
+	options?: ProviderStreamOptions,
+): AssistantMessageEventStream {
+	const provider = resolveApiProvider(model.api);
+	return provider.stream(model, context, withEnvApiKey(model, options) as StreamOptions);
+}
+
+export async function complete<TApi extends Api>(
+	model: Model<TApi>,
+	context: Context,
+	options?: ProviderStreamOptions,
+): Promise<AssistantMessage> {
+	const s = stream(model, context, options);
+	return s.result();
+}
+
+export function streamSimple<TApi extends Api>(
+	model: Model<TApi>,
+	context: Context,
+	options?: SimpleStreamOptions,
+): AssistantMessageEventStream {
+	const provider = resolveApiProvider(model.api);
+	return provider.streamSimple(model, context, withEnvApiKey(model, options));
+}
+
+export async function completeSimple<TApi extends Api>(
+	model: Model<TApi>,
+	context: Context,
+	options?: SimpleStreamOptions,
+): Promise<AssistantMessage> {
+	const s = streamSimple(model, context, options);
+	return s.result();
+}

packages/ai/src/images-models.ts

@@ -0,0 +1,262 @@
+import { defaultProviderAuthContext as defaultAuthContext } from "./auth/context.ts";
+import { InMemoryCredentialStore } from "./auth/credential-store.ts";
+import { ModelsError, resolveProviderAuth } from "./auth/resolve.ts";
+import type { AuthContext, AuthResult, CredentialStore, ProviderAuth } from "./auth/types.ts";
+import type { CreateModelsOptions } from "./models.ts";
+import type { AssistantImages, ImagesApi, ImagesContext, ImagesModel, ImagesOptions, ProviderImages } from "./types.ts";
+
+/**
+ * An image-generation provider: the image-side counterpart of `Provider`.
+ * Owns id/name metadata, auth, model listing, and generation behavior.
+ */
+export interface ImagesProvider {
+	readonly id: string;
+	readonly name: string;
+
+	/**
+	 * Required: at least one of `apiKey`/`oauth`. Same semantics as chat
+	 * providers; `ImagesModels.getAuth()` returns undefined when the provider
+	 * is unconfigured.
+	 */
+	readonly auth: ProviderAuth;
+
+	/**
+	 * Current known models, sync. Static providers return their catalog;
+	 * dynamic providers return the list as of the last `refreshModels()`
+	 * (empty before the first). Must not throw; `ImagesModels` treats a
+	 * throwing implementation as having no models.
+	 */
+	getModels(): readonly ImagesModel<ImagesApi>[];
+
+	/**
+	 * Dynamic providers only: fetch and update the model list. May reject
+	 * (network); on rejection the model list stays at its last-known state
+	 * and a later call retries.
+	 */
+	refreshModels?(): Promise<void>;
+
+	generateImages(
+		model: ImagesModel<ImagesApi>,
+		context: ImagesContext,
+		options?: ImagesOptions,
+	): Promise<AssistantImages>;
+}
+
+/**
+ * Runtime collection of image-generation providers plus auth application and
+ * generation convenience: the image-side counterpart of `Models`.
+ */
+export interface ImagesModels {
+	getProviders(): readonly ImagesProvider[];
+	getProvider(id: string): ImagesProvider | undefined;
+
+	/**
+	 * Sync read of last-known models from one provider or all providers.
+	 * Best-effort: a provider whose `getModels()` throws yields no models.
+	 */
+	getModels(provider?: string): readonly ImagesModel<ImagesApi>[];
+
+	/** Sync runtime model lookup against last-known lists. */
+	getModel(provider: string, id: string): ImagesModel<ImagesApi> | undefined;
+
+	/**
+	 * Ask dynamic providers to re-fetch their model lists. With a provider id,
+	 * rejects with `ModelsError` ("model_source") on that provider's fetch
+	 * failure; without one, refreshes all providers concurrently best-effort.
+	 * Static providers (no `refreshModels`) are no-ops.
+	 */
+	refresh(provider?: string): Promise<void>;
+
+	/**
+	 * Resolve request auth for an image model. Same contract as
+	 * `Models.getAuth()`: undefined when unknown/unconfigured, rejects with
+	 * `ModelsError` ("oauth"/"auth") on real failures.
+	 */
+	getAuth(model: ImagesModel<ImagesApi>): Promise<AuthResult | undefined>;
+
+	/**
+	 * Generate images through the owning provider with auth resolved and
+	 * merged (explicit options win per field). Never rejects; failures are
+	 * returned as an `AssistantImages` with `stopReason: "error"`.
+	 */
+	generateImages(
+		model: ImagesModel<ImagesApi>,
+		context: ImagesContext,
+		options?: ImagesOptions,
+	): Promise<AssistantImages>;
+}
+
+export interface MutableImagesModels extends ImagesModels {
+	/** Upsert/replace by provider.id. Provider ids are unique. */
+	setProvider(provider: ImagesProvider): void;
+	deleteProvider(id: string): void;
+	clearProviders(): void;
+}
+
+class ImagesModelsImpl implements MutableImagesModels {
+	private providers = new Map<string, ImagesProvider>();
+	private credentials: CredentialStore;
+	private authContext: AuthContext;
+
+	constructor(options?: CreateModelsOptions) {
+		this.credentials = options?.credentials ?? new InMemoryCredentialStore();
+		this.authContext = options?.authContext ?? defaultAuthContext();
+	}
+
+	setProvider(provider: ImagesProvider): void {
+		this.providers.set(provider.id, provider);
+	}
+
+	deleteProvider(id: string): void {
+		this.providers.delete(id);
+	}
+
+	clearProviders(): void {
+		this.providers.clear();
+	}
+
+	getProviders(): readonly ImagesProvider[] {
+		return Array.from(this.providers.values());
+	}
+
+	getProvider(id: string): ImagesProvider | undefined {
+		return this.providers.get(id);
+	}
+
+	getModels(provider?: string): readonly ImagesModel<ImagesApi>[] {
+		if (provider !== undefined) {
+			const entry = this.providers.get(provider);
+			if (!entry) return [];
+			try {
+				return entry.getModels();
+			} catch {
+				return [];
+			}
+		}
+
+		const models: ImagesModel<ImagesApi>[] = [];
+		for (const entry of this.providers.values()) {
+			try {
+				models.push(...entry.getModels());
+			} catch {
+				// Best-effort: ill-behaved providers yield no models.
+			}
+		}
+		return models;
+	}
+
+	getModel(provider: string, id: string): ImagesModel<ImagesApi> | undefined {
+		return this.getModels(provider).find((model) => model.id === id);
+	}
+
+	async refresh(provider?: string): Promise<void> {
+		if (provider !== undefined) {
+			const entry = this.providers.get(provider);
+			if (!entry?.refreshModels) return;
+			try {
+				await entry.refreshModels();
+			} catch (error) {
+				if (error instanceof ModelsError) throw error;
+				throw new ModelsError("model_source", `Model refresh failed for ${provider}`, { cause: error });
+			}
+			return;
+		}
+
+		// Cannot reject: the async mapper turns even sync throws from ill-behaved
+		// providers into rejections, and allSettled captures all of them.
+		await Promise.allSettled(Array.from(this.providers.values(), async (entry) => entry.refreshModels?.()));
+	}
+
+	async getAuth(model: ImagesModel<ImagesApi>): Promise<AuthResult | undefined> {
+		const provider = this.providers.get(model.provider);
+		if (!provider) return undefined;
+		return resolveProviderAuth(provider, model, this.credentials, this.authContext);
+	}
+
+	async generateImages(
+		model: ImagesModel<ImagesApi>,
+		context: ImagesContext,
+		options?: ImagesOptions,
+	): Promise<AssistantImages> {
+		try {
+			const provider = this.providers.get(model.provider);
+			if (!provider) {
+				throw new ModelsError("provider", `Unknown provider: ${model.provider}`);
+			}
+
+			const resolution = await this.getAuth(model);
+			const auth = resolution?.auth;
+			if (!auth) {
+				return provider.generateImages(model, context, options);
+			}
+
+			const requestModel = auth.baseUrl ? { ...model, baseUrl: auth.baseUrl } : model;
+
+			// Explicit request options win per-field; headers merge per header.
+			const apiKey = options?.apiKey ?? auth.apiKey;
+			const headers = auth.headers || options?.headers ? { ...auth.headers, ...options?.headers } : undefined;
+
+			return await provider.generateImages(requestModel, context, { ...options, apiKey, headers });
+		} catch (error) {
+			return {
+				api: model.api,
+				provider: model.provider,
+				model: model.id,
+				output: [],
+				stopReason: "error",
+				errorMessage: error instanceof Error ? error.message : String(error),
+				timestamp: Date.now(),
+			};
+		}
+	}
+}
+
+export function createImagesModels(options?: CreateModelsOptions): MutableImagesModels {
+	return new ImagesModelsImpl(options);
+}
+
+export interface CreateImagesProviderOptions {
+	id: string;
+	/** Display name. Default: `id`. */
+	name?: string;
+	/** Required — every provider has auth semantics, even ambient/keyless ones. */
+	auth: ProviderAuth;
+	/** Initial model list (empty for purely dynamic providers). */
+	models: readonly ImagesModel<ImagesApi>[];
+	/**
+	 * Dynamic providers: fetch the current list. Stored on success; concurrent
+	 * calls share one in-flight fetch. May reject: the stored list then stays
+	 * at its last-known state, the rejection propagates to the caller of
+	 * `refreshModels()` (wrapped as ModelsError "model_source" by
+	 * `ImagesModels.refresh(provider)`), and a later call retries.
+	 */
+	refreshModels?: () => Promise<readonly ImagesModel<ImagesApi>[]>;
+	api: ProviderImages;
+}
+
+/** Builds an image-generation provider from parts. */
+export function createImagesProvider(input: CreateImagesProviderOptions): ImagesProvider {
+	let models = input.models;
+	let inflightRefresh: Promise<void> | undefined;
+	const refreshModels = input.refreshModels;
+
+	return {
+		id: input.id,
+		name: input.name ?? input.id,
+		auth: input.auth,
+		getModels: () => models,
+		refreshModels: refreshModels
+			? () => {
+					inflightRefresh ??= (async () => {
+						try {
+							models = await refreshModels();
+						} finally {
+							inflightRefresh = undefined;
+						}
+					})();
+					return inflightRefresh;
+				}
+			: undefined,
+		generateImages: (model, context, options) => input.api.generateImages(model, context, options),
+	};
+}

packages/ai/src/index.ts

@@ -1,30 +1,30 @@
 export type { Static, TSchema } from "typebox";
 export { Type } from "typebox";
 
-export * from "./api-registry.ts";
-export * from "./env-api-keys.ts";
-export * from "./image-models.ts";
-export * from "./images.ts";
-export * from "./images-api-registry.ts";
+// Core only, side-effect free: no generated catalogs, no provider factories,
+// no api-registry, no OAuth implementations, no compat. Provider factories
+// live under "@earendil-works/pi-ai/providers/*", API implementations under
+// "@earendil-works/pi-ai/api/*", the old global API under
+// "@earendil-works/pi-ai/compat".
+export type { AnthropicEffort, AnthropicOptions, AnthropicThinkingDisplay } from "./api/anthropic-messages.ts";
+export type { AzureOpenAIResponsesOptions } from "./api/azure-openai-responses.ts";
+export type { BedrockOptions, BedrockThinkingDisplay } from "./api/bedrock-converse-stream.ts";
+export type { GoogleOptions } from "./api/google-generative-ai.ts";
+export type { GoogleThinkingLevel } from "./api/google-shared.ts";
+export type { GoogleVertexOptions } from "./api/google-vertex.ts";
+export * from "./api/lazy.ts";
+export type { MistralOptions } from "./api/mistral-conversations.ts";
+export type { OpenAICodexResponsesOptions, OpenAICodexWebSocketDebugStats } from "./api/openai-codex-responses.ts";
+export type { OpenAICompletionsOptions } from "./api/openai-completions.ts";
+export type { OpenAIResponsesOptions } from "./api/openai-responses.ts";
+export * from "./auth/context.ts";
+export * from "./auth/credential-store.ts";
+export * from "./auth/helpers.ts";
+export * from "./auth/types.ts";
+export * from "./images-models.ts";
 export * from "./models.ts";
-export type { BedrockOptions, BedrockThinkingDisplay } from "./providers/amazon-bedrock.ts";
-export type { AnthropicEffort, AnthropicOptions, AnthropicThinkingDisplay } from "./providers/anthropic.ts";
-export type { AzureOpenAIResponsesOptions } from "./providers/azure-openai-responses.ts";
 export * from "./providers/faux.ts";
-export type { GoogleOptions } from "./providers/google.ts";
-export type { GoogleThinkingLevel } from "./providers/google-shared.ts";
-export type { GoogleVertexOptions } from "./providers/google-vertex.ts";
-export * from "./providers/images/register-builtins.ts";
-export type { MistralOptions } from "./providers/mistral.ts";
-export type {
-	OpenAICodexResponsesOptions,
-	OpenAICodexWebSocketDebugStats,
-} from "./providers/openai-codex-responses.ts";
-export type { OpenAICompletionsOptions } from "./providers/openai-completions.ts";
-export type { OpenAIResponsesOptions } from "./providers/openai-responses.ts";
-export * from "./providers/register-builtins.ts";
 export * from "./session-resources.ts";
-export * from "./stream.ts";
 export * from "./types.ts";
 export * from "./utils/diagnostics.ts";
 export * from "./utils/event-stream.ts";

packages/ai/src/models.ts

@@ -1,39 +1,374 @@
-import { MODELS } from "./models.generated.ts";
-import type { Api, KnownProvider, Model, ModelThinkingLevel, Usage } from "./types.ts";
+import { lazyStream } from "./api/lazy.ts";
+import { defaultProviderAuthContext as defaultAuthContext } from "./auth/context.ts";
+import { InMemoryCredentialStore } from "./auth/credential-store.ts";
+import { ModelsError, resolveProviderAuth } from "./auth/resolve.ts";
+import type { AuthContext, AuthResult, CredentialStore, ProviderAuth } from "./auth/types.ts";
+import type {
+	Api,
+	ApiStreamOptions,
+	AssistantMessage,
+	AssistantMessageEventStream,
+	Context,
+	Model,
+	ModelThinkingLevel,
+	ProviderStreams,
+	SimpleStreamOptions,
+	StreamOptions,
+	Usage,
+} from "./types.ts";
 
-const modelRegistry: Map<string, Map<string, Model<Api>>> = new Map();
+export { type AuthModel, ModelsError, type ModelsErrorCode } from "./auth/resolve.ts";
 
-// Initialize registry from MODELS on module load
-for (const [provider, models] of Object.entries(MODELS)) {
-	const providerModels = new Map<string, Model<Api>>();
-	for (const [id, model] of Object.entries(models)) {
-		providerModels.set(id, model as Model<Api>);
+/**
+ * A provider is the concrete runtime unit. It owns id/name/base metadata,
+ * auth methods, model listing, and stream behavior.
+ *
+ * `TApi` lets concrete provider factories declare which APIs their models
+ * use (e.g. `openaiProvider(): Provider<"openai-responses" | "openai-completions">`),
+ * giving typed model lists to direct factory users. Inside a `Models`
+ * collection providers are held as `Provider<Api>`.
+ */
+export interface Provider<TApi extends Api = Api> {
+	readonly id: string;
+	readonly name: string;
+
+	readonly baseUrl?: string;
+	readonly headers?: Record<string, string>;
+
+	/**
+	 * Required: at least one of `apiKey`/`oauth`. Every provider has auth
+	 * semantics — even providers with only ambient credentials (env vars, AWS
+	 * profiles, ADC files) and keyless local servers provide `apiKey` auth
+	 * whose `resolve()` reports whether the provider is configured.
+	 * `Models.getAuth()` returns undefined when the provider is unconfigured.
+	 */
+	readonly auth: ProviderAuth;
+
+	/**
+	 * Current known models, sync. Static providers return their catalog;
+	 * dynamic providers return the list as of the last `refreshModels()`
+	 * (empty before the first). Must not throw; `Models` treats a throwing
+	 * implementation as having no models.
+	 */
+	getModels(): readonly Model<TApi>[];
+
+	/**
+	 * Dynamic providers only: fetch and update the model list. Side-effect-free
+	 * discovery (no loading/downloading); provider-specific model lifecycle
+	 * belongs in app commands. Concurrent calls share one in-flight fetch.
+	 * May reject (network); on rejection the model list stays at its last-known
+	 * state and a later call retries.
+	 */
+	refreshModels?(): Promise<void>;
+
+	stream<T extends TApi>(
+		model: Model<T>,
+		context: Context,
+		options?: ApiStreamOptions<T>,
+	): AssistantMessageEventStream;
+
+	streamSimple(model: Model<TApi>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream;
+}
+
+/**
+ * Runtime collection of providers plus auth application and stream
+ * convenience. Providers own stream behavior; `Models` resolves auth and
+ * delegates each request to the provider that owns the model.
+ */
+export interface Models {
+	getProviders(): readonly Provider[];
+	getProvider(id: string): Provider | undefined;
+
+	/**
+	 * Sync read of last-known models from one provider or all providers.
+	 * Best-effort: a provider whose `getModels()` throws yields no models.
+	 */
+	getModels(provider?: string): readonly Model<Api>[];
+
+	/**
+	 * Sync runtime model lookup against last-known lists. Dynamic model lists
+	 * are typed as `Model<Api>`; narrow with the `hasApi()` type guard.
+	 */
+	getModel(provider: string, id: string): Model<Api> | undefined;
+
+	/**
+	 * Ask dynamic providers to re-fetch their model lists. With a provider id,
+	 * rejects with `ModelsError` ("model_source") on that provider's fetch
+	 * failure; without one, refreshes all providers concurrently best-effort.
+	 * Static providers (no `refreshModels`) are no-ops.
+	 */
+	refresh(provider?: string): Promise<void>;
+
+	/**
+	 * Resolve request auth for a model. Includes a source label for status UI.
+	 * Resolves `undefined` when the provider is unknown or unconfigured.
+	 * Rejects with `ModelsError`: code "oauth" when a token refresh fails (the
+	 * stored credential is preserved for retry; re-login fixes it), code "auth"
+	 * when api-key resolution or the credential store fails. Request paths
+	 * surface rejections as stream errors; status/availability UIs catch them
+	 * and render "needs re-login" instead of treating them as unconfigured.
+	 */
+	getAuth(model: Model<Api>): Promise<AuthResult | undefined>;
+
+	stream<TApi extends Api>(
+		model: Model<TApi>,
+		context: Context,
+		options?: ApiStreamOptions<TApi>,
+	): AssistantMessageEventStream;
+
+	complete<TApi extends Api>(
+		model: Model<TApi>,
+		context: Context,
+		options?: ApiStreamOptions<TApi>,
+	): Promise<AssistantMessage>;
+
+	streamSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream;
+	completeSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): Promise<AssistantMessage>;
+}
+
+export interface MutableModels extends Models {
+	/** Upsert/replace by provider.id. Provider ids are unique. */
+	setProvider(provider: Provider): void;
+	deleteProvider(id: string): void;
+	clearProviders(): void;
+}
+
+export interface CreateModelsOptions {
+	credentials?: CredentialStore;
+	authContext?: AuthContext;
+}
+
+class ModelsImpl implements MutableModels {
+	private providers = new Map<string, Provider>();
+	private credentials: CredentialStore;
+	private authContext: AuthContext;
+
+	constructor(options?: CreateModelsOptions) {
+		this.credentials = options?.credentials ?? new InMemoryCredentialStore();
+		this.authContext = options?.authContext ?? defaultAuthContext();
+	}
+
+	setProvider(provider: Provider): void {
+		this.providers.set(provider.id, provider);
+	}
+
+	deleteProvider(id: string): void {
+		this.providers.delete(id);
+	}
+
+	clearProviders(): void {
+		this.providers.clear();
+	}
+
+	getProviders(): readonly Provider[] {
+		return Array.from(this.providers.values());
+	}
+
+	getProvider(id: string): Provider | undefined {
+		return this.providers.get(id);
+	}
+
+	getModels(provider?: string): readonly Model<Api>[] {
+		if (provider !== undefined) {
+			const entry = this.providers.get(provider);
+			if (!entry) return [];
+			try {
+				return entry.getModels();
+			} catch {
+				return [];
+			}
+		}
+
+		const models: Model<Api>[] = [];
+		for (const entry of this.providers.values()) {
+			try {
+				models.push(...entry.getModels());
+			} catch {
+				// Best-effort: ill-behaved providers yield no models.
+			}
+		}
+		return models;
+	}
+
+	getModel(provider: string, id: string): Model<Api> | undefined {
+		return this.getModels(provider).find((model) => model.id === id);
+	}
+
+	async refresh(provider?: string): Promise<void> {
+		if (provider !== undefined) {
+			const entry = this.providers.get(provider);
+			if (!entry?.refreshModels) return;
+			try {
+				await entry.refreshModels();
+			} catch (error) {
+				if (error instanceof ModelsError) throw error;
+				throw new ModelsError("model_source", `Model refresh failed for ${provider}`, { cause: error });
+			}
+			return;
+		}
+
+		// Cannot reject: the async mapper turns even sync throws from ill-behaved
+		// providers into rejections, and allSettled captures all of them.
+		await Promise.allSettled(Array.from(this.providers.values(), async (entry) => entry.refreshModels?.()));
+	}
+
+	async getAuth(model: Model<Api>): Promise<AuthResult | undefined> {
+		const provider = this.providers.get(model.provider);
+		if (!provider) return undefined;
+		return resolveProviderAuth(provider, model, this.credentials, this.authContext);
+	}
+
+	private requireProvider(model: Model<Api>): Provider {
+		const provider = this.providers.get(model.provider);
+		if (!provider) {
+			throw new ModelsError("provider", `Unknown provider: ${model.provider}`);
+		}
+		return provider;
+	}
+
+	private async applyAuth<TOptions extends StreamOptions>(
+		model: Model<Api>,
+		options: TOptions | undefined,
+	): Promise<{ requestModel: Model<Api>; requestOptions: TOptions | undefined }> {
+		const resolution = await this.getAuth(model);
+		const auth = resolution?.auth;
+		if (!auth) return { requestModel: model, requestOptions: options };
+
+		const requestModel = auth.baseUrl ? { ...model, baseUrl: auth.baseUrl } : model;
+
+		// Explicit request options win per-field; headers merge per header.
+		const apiKey = options?.apiKey ?? auth.apiKey;
+		const headers = auth.headers || options?.headers ? { ...auth.headers, ...options?.headers } : undefined;
+		const requestOptions = { ...options, apiKey, headers } as TOptions;
+
+		return { requestModel, requestOptions };
+	}
+
+	stream<TApi extends Api>(
+		model: Model<TApi>,
+		context: Context,
+		options?: ApiStreamOptions<TApi>,
+	): AssistantMessageEventStream {
+		return lazyStream(model, async () => {
+			const provider = this.requireProvider(model);
+			const { requestModel, requestOptions } = await this.applyAuth(model, options as StreamOptions | undefined);
+			return provider.stream(requestModel as Model<TApi>, context, requestOptions as ApiStreamOptions<TApi>);
+		});
+	}
+
+	async complete<TApi extends Api>(
+		model: Model<TApi>,
+		context: Context,
+		options?: ApiStreamOptions<TApi>,
+	): Promise<AssistantMessage> {
+		return this.stream(model, context, options).result();
+	}
+
+	streamSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): AssistantMessageEventStream {
+		return lazyStream(model, async () => {
+			const provider = this.requireProvider(model);
+			const { requestModel, requestOptions } = await this.applyAuth(model, options);
+			return provider.streamSimple(requestModel, context, requestOptions);
+		});
+	}
+
+	async completeSimple(model: Model<Api>, context: Context, options?: SimpleStreamOptions): Promise<AssistantMessage> {
+		return this.streamSimple(model, context, options).result();
 	}
-	modelRegistry.set(provider, providerModels);
 }
 
-type ModelApi<
-	TProvider extends KnownProvider,
-	TModelId extends keyof (typeof MODELS)[TProvider],
-> = (typeof MODELS)[TProvider][TModelId] extends { api: infer TApi } ? (TApi extends Api ? TApi : never) : never;
-
-export function getModel<TProvider extends KnownProvider, TModelId extends keyof (typeof MODELS)[TProvider]>(
-	provider: TProvider,
-	modelId: TModelId,
-): Model<ModelApi<TProvider, TModelId>> {
-	const providerModels = modelRegistry.get(provider);
-	return providerModels?.get(modelId as string) as Model<ModelApi<TProvider, TModelId>>;
+export function createModels(options?: CreateModelsOptions): MutableModels {
+	return new ModelsImpl(options);
 }
 
-export function getProviders(): KnownProvider[] {
-	return Array.from(modelRegistry.keys()) as KnownProvider[];
+export interface CreateProviderOptions<TApi extends Api = Api> {
+	id: string;
+	/** Display name. Default: `id`. */
+	name?: string;
+	baseUrl?: string;
+	headers?: Record<string, string>;
+	/** Required — every provider has auth semantics, even ambient/keyless ones. */
+	auth: ProviderAuth;
+	/** Initial model list (empty for purely dynamic providers). */
+	models: readonly Model<TApi>[];
+	/**
+	 * Dynamic providers: fetch the current list. Stored on success; concurrent
+	 * calls share one in-flight fetch. May reject: the stored list then stays
+	 * at its last-known state, the rejection propagates to the caller of
+	 * `refreshModels()` (wrapped as ModelsError "model_source" by
+	 * `Models.refresh(provider)`), and a later call retries.
+	 */
+	refreshModels?: () => Promise<readonly Model<TApi>[]>;
+	/** Single implementation, or map keyed by `model.api` for mixed-API providers. */
+	api: ProviderStreams | Partial<Record<TApi, ProviderStreams>>;
 }
 
-export function getModels<TProvider extends KnownProvider>(
-	provider: TProvider,
-): Model<ModelApi<TProvider, keyof (typeof MODELS)[TProvider]>>[] {
-	const models = modelRegistry.get(provider);
-	return models ? (Array.from(models.values()) as Model<ModelApi<TProvider, keyof (typeof MODELS)[TProvider]>>[]) : [];
+/**
+ * Builds a provider from parts. Built-in provider factories and models.json
+ * custom providers both go through this. A single `api` streams all models;
+ * an `api` map dispatches on `model.api`, and a model whose api has no entry
+ * produces a stream error.
+ */
+export function createProvider<TApi extends Api = Api>(input: CreateProviderOptions<TApi>): Provider<TApi> {
+	let models = input.models;
+	let inflightRefresh: Promise<void> | undefined;
+	const refreshModels = input.refreshModels;
+	const single =
+		typeof (input.api as ProviderStreams).stream === "function" ? (input.api as ProviderStreams) : undefined;
+	const byApi = single ? undefined : (input.api as Partial<Record<string, ProviderStreams>>);
+
+	const apiFor = (model: Model<Api>): ProviderStreams | undefined => single ?? byApi?.[model.api];
+
+	const dispatch = (
+		model: Model<Api>,
+		run: (streams: ProviderStreams) => AssistantMessageEventStream,
+	): AssistantMessageEventStream => {
+		const streams = apiFor(model);
+		if (!streams) {
+			return lazyStream(model, async () => {
+				throw new ModelsError("stream", `Provider ${input.id} has no API implementation for "${model.api}"`);
+			});
+		}
+		return run(streams);
+	};
+
+	return {
+		id: input.id,
+		name: input.name ?? input.id,
+		baseUrl: input.baseUrl,
+		headers: input.headers,
+		auth: input.auth,
+		getModels: () => models,
+		refreshModels: refreshModels
+			? () => {
+					inflightRefresh ??= (async () => {
+						try {
+							models = await refreshModels();
+						} finally {
+							inflightRefresh = undefined;
+						}
+					})();
+					return inflightRefresh;
+				}
+			: undefined,
+		stream: (model, context, options) => dispatch(model, (streams) => streams.stream(model, context, options)),
+		streamSimple: (model, context, options) =>
+			dispatch(model, (streams) => streams.streamSimple(model, context, options)),
+	};
+}
+
+/**
+ * Runtime-checked narrowing for dynamically looked-up models:
+ *
+ * ```ts
+ * const model = models.getModel("anthropic", "claude-opus-4-7");
+ * if (model && hasApi(model, "anthropic-messages")) {
+ *   // model: Model<"anthropic-messages">, stream options fully typed
+ * }
+ * ```
+ */
+export function hasApi<TApi extends Api>(model: Model<Api>, api: TApi): model is Model<TApi> {
+	return model.api === api;
 }
 
 export function calculateCost<TApi extends Api>(model: Model<TApi>, usage: Usage): Usage["cost"] {

packages/ai/src/providers/all.ts

@@ -0,0 +1,131 @@
+import { createImagesModels, type ImagesProvider, type MutableImagesModels } from "../images-models.ts";
+import { MODELS } from "../models.generated.ts";
+import { type CreateModelsOptions, createModels, type MutableModels, type Provider } from "../models.ts";
+import type { Api, KnownProvider, Model } from "../types.ts";
+import { amazonBedrockProvider } from "./amazon-bedrock.ts";
+import { antLingProvider } from "./ant-ling.ts";
+import { anthropicProvider } from "./anthropic.ts";
+import { azureOpenAIResponsesProvider } from "./azure-openai-responses.ts";
+import { cerebrasProvider } from "./cerebras.ts";
+import { cloudflareAIGatewayProvider } from "./cloudflare-ai-gateway.ts";
+import { cloudflareWorkersAIProvider } from "./cloudflare-workers-ai.ts";
+import { deepseekProvider } from "./deepseek.ts";
+import { fireworksProvider } from "./fireworks.ts";
+import { githubCopilotProvider } from "./github-copilot.ts";
+import { googleProvider } from "./google.ts";
+import { googleVertexProvider } from "./google-vertex.ts";
+import { groqProvider } from "./groq.ts";
+import { huggingfaceProvider } from "./huggingface.ts";
+import { kimiCodingProvider } from "./kimi-coding.ts";
+import { minimaxProvider } from "./minimax.ts";
+import { minimaxCnProvider } from "./minimax-cn.ts";
+import { mistralProvider } from "./mistral.ts";
+import { moonshotaiProvider } from "./moonshotai.ts";
+import { moonshotaiCnProvider } from "./moonshotai-cn.ts";
+import { nvidiaProvider } from "./nvidia.ts";
+import { openaiProvider } from "./openai.ts";
+import { openaiCodexProvider } from "./openai-codex.ts";
+import { opencodeProvider } from "./opencode.ts";
+import { opencodeGoProvider } from "./opencode-go.ts";
+import { openrouterProvider } from "./openrouter.ts";
+import { openrouterImagesProvider } from "./openrouter-images.ts";
+import { togetherProvider } from "./together.ts";
+import { vercelAIGatewayProvider } from "./vercel-ai-gateway.ts";
+import { xaiProvider } from "./xai.ts";
+import { xiaomiProvider } from "./xiaomi.ts";
+import { xiaomiTokenPlanAmsProvider } from "./xiaomi-token-plan-ams.ts";
+import { xiaomiTokenPlanCnProvider } from "./xiaomi-token-plan-cn.ts";
+import { xiaomiTokenPlanSgpProvider } from "./xiaomi-token-plan-sgp.ts";
+import { zaiProvider } from "./zai.ts";
+import { zaiCodingCnProvider } from "./zai-coding-cn.ts";
+
+type BuiltinModelApi<
+	TProvider extends KnownProvider,
+	TModelId extends keyof (typeof MODELS)[TProvider],
+> = (typeof MODELS)[TProvider][TModelId] extends { api: infer TApi } ? (TApi extends Api ? TApi : never) : never;
+
+/** Typed read of the generated built-in catalog. */
+export function getBuiltinModel<TProvider extends KnownProvider, TModelId extends keyof (typeof MODELS)[TProvider]>(
+	provider: TProvider,
+	modelId: TModelId,
+): Model<BuiltinModelApi<TProvider, TModelId>> {
+	const models = MODELS[provider] as Record<string, Model<Api>> | undefined;
+	return models?.[modelId as string] as Model<BuiltinModelApi<TProvider, TModelId>>;
+}
+
+export function getBuiltinProviders(): KnownProvider[] {
+	return Object.keys(MODELS) as KnownProvider[];
+}
+
+export function getBuiltinModels<TProvider extends KnownProvider>(
+	provider: TProvider,
+): Model<BuiltinModelApi<TProvider, keyof (typeof MODELS)[TProvider]>>[] {
+	const models = MODELS[provider] as Record<string, Model<Api>> | undefined;
+	return models
+		? (Object.values(models) as Model<BuiltinModelApi<TProvider, keyof (typeof MODELS)[TProvider]>>[])
+		: [];
+}
+
+/** All built-in providers, freshly constructed. */
+export function builtinProviders(): Provider[] {
+	return [
+		amazonBedrockProvider(),
+		antLingProvider(),
+		anthropicProvider(),
+		azureOpenAIResponsesProvider(),
+		cerebrasProvider(),
+		cloudflareAIGatewayProvider(),
+		cloudflareWorkersAIProvider(),
+		deepseekProvider(),
+		fireworksProvider(),
+		githubCopilotProvider(),
+		googleProvider(),
+		googleVertexProvider(),
+		groqProvider(),
+		huggingfaceProvider(),
+		kimiCodingProvider(),
+		minimaxProvider(),
+		minimaxCnProvider(),
+		mistralProvider(),
+		moonshotaiProvider(),
+		moonshotaiCnProvider(),
+		nvidiaProvider(),
+		openaiProvider(),
+		openaiCodexProvider(),
+		opencodeProvider(),
+		opencodeGoProvider(),
+		openrouterProvider(),
+		togetherProvider(),
+		vercelAIGatewayProvider(),
+		xaiProvider(),
+		xiaomiProvider(),
+		xiaomiTokenPlanAmsProvider(),
+		xiaomiTokenPlanCnProvider(),
+		xiaomiTokenPlanSgpProvider(),
+		zaiProvider(),
+		zaiCodingCnProvider(),
+	];
+}
+
+/** A `Models` collection with every built-in provider registered. */
+export function builtinModels(options?: CreateModelsOptions): MutableModels {
+	const models = createModels(options);
+	for (const provider of builtinProviders()) {
+		models.setProvider(provider);
+	}
+	return models;
+}
+
+/** All built-in image-generation providers, freshly constructed. */
+export function builtinImagesProviders(): ImagesProvider[] {
+	return [openrouterImagesProvider()];
+}
+
+/** An `ImagesModels` collection with every built-in image-generation provider registered. */
+export function builtinImagesModels(options?: CreateModelsOptions): MutableImagesModels {
+	const models = createImagesModels(options);
+	for (const provider of builtinImagesProviders()) {
+		models.setProvider(provider);
+	}
+	return models;
+}

packages/ai/src/providers/ant-ling.models.ts

@@ -0,0 +1,62 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const ANT_LING_MODELS = {
+	"Ling-2.6-1T": {
+		id: "Ling-2.6-1T",
+		name: "Ling 2.6 1T",
+		api: "openai-completions",
+		provider: "ant-ling",
+		baseUrl: "https://api.ant-ling.com/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsLongCacheRetention":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.06,
+			output: 0.25,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"Ling-2.6-flash": {
+		id: "Ling-2.6-flash",
+		name: "Ling 2.6 Flash",
+		api: "openai-completions",
+		provider: "ant-ling",
+		baseUrl: "https://api.ant-ling.com/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsLongCacheRetention":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.01,
+			output: 0.02,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"Ring-2.6-1T": {
+		id: "Ring-2.6-1T",
+		name: "Ring 2.6 1T",
+		api: "openai-completions",
+		provider: "ant-ling",
+		baseUrl: "https://api.ant-ling.com/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsLongCacheRetention":false,"thinkingFormat":"ant-ling"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":null,"medium":null,"high":"high","xhigh":"xhigh"},
+		input: ["text"],
+		cost: {
+			input: 0.06,
+			output: 0.25,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/ant-ling.ts

@@ -0,0 +1,15 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { ANT_LING_MODELS } from "./ant-ling.models.ts";
+
+export function antLingProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "ant-ling",
+		name: "Ant Ling",
+		baseUrl: "https://api.ant-ling.com/v1",
+		auth: { apiKey: envApiKeyAuth("Ant Ling API key", ["ANT_LING_API_KEY"]) },
+		models: Object.values(ANT_LING_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/anthropic.models.ts

@@ -0,0 +1,441 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const ANTHROPIC_MODELS = {
+	"claude-3-5-haiku-20241022": {
+		id: "claude-3-5-haiku-20241022",
+		name: "Claude Haiku 3.5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.8,
+			output: 4,
+			cacheRead: 0.08,
+			cacheWrite: 1,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-5-haiku-latest": {
+		id: "claude-3-5-haiku-latest",
+		name: "Claude Haiku 3.5 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.8,
+			output: 4,
+			cacheRead: 0.08,
+			cacheWrite: 1,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-5-sonnet-20240620": {
+		id: "claude-3-5-sonnet-20240620",
+		name: "Claude Sonnet 3.5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-5-sonnet-20241022": {
+		id: "claude-3-5-sonnet-20241022",
+		name: "Claude Sonnet 3.5 v2",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-7-sonnet-20250219": {
+		id: "claude-3-7-sonnet-20250219",
+		name: "Claude Sonnet 3.7",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-haiku-20240307": {
+		id: "claude-3-haiku-20240307",
+		name: "Claude Haiku 3",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.25,
+			cacheRead: 0.03,
+			cacheWrite: 0.3,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-opus-20240229": {
+		id: "claude-3-opus-20240229",
+		name: "Claude Opus 3",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-sonnet-20240229": {
+		id: "claude-3-sonnet-20240229",
+		name: "Claude Sonnet 3",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 0.3,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-fable-5": {
+		id: "claude-fable-5",
+		name: "Claude Fable 5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 50,
+			cacheRead: 1,
+			cacheWrite: 12.5,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-haiku-4-5": {
+		id: "claude-haiku-4-5",
+		name: "Claude Haiku 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1,
+			output: 5,
+			cacheRead: 0.1,
+			cacheWrite: 1.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-haiku-4-5-20251001": {
+		id: "claude-haiku-4-5-20251001",
+		name: "Claude Haiku 4.5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1,
+			output: 5,
+			cacheRead: 0.1,
+			cacheWrite: 1.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-0": {
+		id: "claude-opus-4-0",
+		name: "Claude Opus 4 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-1": {
+		id: "claude-opus-4-1",
+		name: "Claude Opus 4.1 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-1-20250805": {
+		id: "claude-opus-4-1-20250805",
+		name: "Claude Opus 4.1",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-20250514": {
+		id: "claude-opus-4-20250514",
+		name: "Claude Opus 4",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-5": {
+		id: "claude-opus-4-5",
+		name: "Claude Opus 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-5-20251101": {
+		id: "claude-opus-4-5-20251101",
+		name: "Claude Opus 4.5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-6": {
+		id: "claude-opus-4-6",
+		name: "Claude Opus 4.6",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"max"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-7": {
+		id: "claude-opus-4-7",
+		name: "Claude Opus 4.7",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-8": {
+		id: "claude-opus-4-8",
+		name: "Claude Opus 4.8",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-0": {
+		id: "claude-sonnet-4-0",
+		name: "Claude Sonnet 4 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-20250514": {
+		id: "claude-sonnet-4-20250514",
+		name: "Claude Sonnet 4",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-5": {
+		id: "claude-sonnet-4-5",
+		name: "Claude Sonnet 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-5-20250929": {
+		id: "claude-sonnet-4-5-20250929",
+		name: "Claude Sonnet 4.5",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-6": {
+		id: "claude-sonnet-4-6",
+		name: "Claude Sonnet 4.6",
+		api: "anthropic-messages",
+		provider: "anthropic",
+		baseUrl: "https://api.anthropic.com",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 1000000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+} as const;

packages/ai/src/providers/azure-openai-responses.models.ts

@@ -0,0 +1,745 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const AZURE_OPENAI_RESPONSES_MODELS = {
+	"gpt-4": {
+		id: "gpt-4",
+		name: "GPT-4",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 30,
+			output: 60,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 8192,
+		maxTokens: 8192,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4-turbo": {
+		id: "gpt-4-turbo",
+		name: "GPT-4 Turbo",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 30,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 4096,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4.1": {
+		id: "gpt-4.1",
+		name: "GPT-4.1",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 1047576,
+		maxTokens: 32768,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4.1-mini": {
+		id: "gpt-4.1-mini",
+		name: "GPT-4.1 mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.4,
+			output: 1.6,
+			cacheRead: 0.1,
+			cacheWrite: 0,
+		},
+		contextWindow: 1047576,
+		maxTokens: 32768,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4.1-nano": {
+		id: "gpt-4.1-nano",
+		name: "GPT-4.1 nano",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.4,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1047576,
+		maxTokens: 32768,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4o": {
+		id: "gpt-4o",
+		name: "GPT-4o",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 10,
+			cacheRead: 1.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4o-2024-05-13": {
+		id: "gpt-4o-2024-05-13",
+		name: "GPT-4o (2024-05-13)",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 15,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 4096,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4o-2024-08-06": {
+		id: "gpt-4o-2024-08-06",
+		name: "GPT-4o (2024-08-06)",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 10,
+			cacheRead: 1.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4o-2024-11-20": {
+		id: "gpt-4o-2024-11-20",
+		name: "GPT-4o (2024-11-20)",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 10,
+			cacheRead: 1.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-4o-mini": {
+		id: "gpt-4o-mini",
+		name: "GPT-4o mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.075,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5": {
+		id: "gpt-5",
+		name: "GPT-5",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5-chat-latest": {
+		id: "gpt-5-chat-latest",
+		name: "GPT-5 Chat Latest",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5-codex": {
+		id: "gpt-5-codex",
+		name: "GPT-5-Codex",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5-mini": {
+		id: "gpt-5-mini",
+		name: "GPT-5 Mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 2,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5-nano": {
+		id: "gpt-5-nano",
+		name: "GPT-5 Nano",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.05,
+			output: 0.4,
+			cacheRead: 0.005,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5-pro": {
+		id: "gpt-5-pro",
+		name: "GPT-5 Pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 120,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.1": {
+		id: "gpt-5.1",
+		name: "GPT-5.1",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.1-chat-latest": {
+		id: "gpt-5.1-chat-latest",
+		name: "GPT-5.1 Chat",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.1-codex": {
+		id: "gpt-5.1-codex",
+		name: "GPT-5.1 Codex",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.1-codex-max": {
+		id: "gpt-5.1-codex-max",
+		name: "GPT-5.1 Codex Max",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.1-codex-mini": {
+		id: "gpt-5.1-codex-mini",
+		name: "GPT-5.1 Codex mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 2,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.2": {
+		id: "gpt-5.2",
+		name: "GPT-5.2",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.2-chat-latest": {
+		id: "gpt-5.2-chat-latest",
+		name: "GPT-5.2 Chat",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.2-codex": {
+		id: "gpt-5.2-codex",
+		name: "GPT-5.2 Codex",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.2-pro": {
+		id: "gpt-5.2-pro",
+		name: "GPT-5.2 Pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 21,
+			output: 168,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.3-chat-latest": {
+		id: "gpt-5.3-chat-latest",
+		name: "GPT-5.3 Chat (latest)",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: false,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.3-codex": {
+		id: "gpt-5.3-codex",
+		name: "GPT-5.3 Codex",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.3-codex-spark": {
+		id: "gpt-5.3-codex-spark",
+		name: "GPT-5.3 Codex Spark",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 32000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.4": {
+		id: "gpt-5.4",
+		name: "GPT-5.4",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 15,
+			cacheRead: 0.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.4-mini": {
+		id: "gpt-5.4-mini",
+		name: "GPT-5.4 mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 0.75,
+			output: 4.5,
+			cacheRead: 0.075,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.4-nano": {
+		id: "gpt-5.4-nano",
+		name: "GPT-5.4 nano",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 0.2,
+			output: 1.25,
+			cacheRead: 0.02,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.4-pro": {
+		id: "gpt-5.4-pro",
+		name: "GPT-5.4 Pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 30,
+			output: 180,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.5": {
+		id: "gpt-5.5",
+		name: "GPT-5.5",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 30,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"gpt-5.5-pro": {
+		id: "gpt-5.5-pro",
+		name: "GPT-5.5 Pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh","minimal":null,"low":null},
+		input: ["text", "image"],
+		cost: {
+			input: 30,
+			output: 180,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"azure-openai-responses">,
+	"o1": {
+		id: "o1",
+		name: "o1",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 60,
+			cacheRead: 7.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o1-pro": {
+		id: "o1-pro",
+		name: "o1-pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 150,
+			output: 600,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o3": {
+		id: "o3",
+		name: "o3",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o3-deep-research": {
+		id: "o3-deep-research",
+		name: "o3-deep-research",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 40,
+			cacheRead: 2.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o3-mini": {
+		id: "o3-mini",
+		name: "o3-mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.1,
+			output: 4.4,
+			cacheRead: 0.55,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o3-pro": {
+		id: "o3-pro",
+		name: "o3-pro",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 20,
+			output: 80,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o4-mini": {
+		id: "o4-mini",
+		name: "o4-mini",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.1,
+			output: 4.4,
+			cacheRead: 0.275,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+	"o4-mini-deep-research": {
+		id: "o4-mini-deep-research",
+		name: "o4-mini-deep-research",
+		api: "azure-openai-responses",
+		provider: "azure-openai-responses",
+		baseUrl: "",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"azure-openai-responses">,
+} as const;

packages/ai/src/providers/azure-openai-responses.ts

@@ -1,299 +1,14 @@
-import { AzureOpenAI } from "openai";
-import type { ResponseCreateParamsStreaming } from "openai/resources/responses/responses.js";
-import { clampThinkingLevel } from "../models.ts";
-import type {
-	Api,
-	AssistantMessage,
-	Context,
-	Model,
-	SimpleStreamOptions,
-	StreamFunction,
-	StreamOptions,
-} from "../types.ts";
-import { AssistantMessageEventStream } from "../utils/event-stream.ts";
-import { headersToRecord } from "../utils/headers.ts";
-import { getProviderEnvValue } from "../utils/provider-env.ts";
-import { clampOpenAIPromptCacheKey } from "./openai-prompt-cache.ts";
-import { convertResponsesMessages, convertResponsesTools, processResponsesStream } from "./openai-responses-shared.ts";
-import { buildBaseOptions } from "./simple-options.ts";
+import { azureOpenAIResponsesApi } from "../api/azure-openai-responses.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { AZURE_OPENAI_RESPONSES_MODELS } from "./azure-openai-responses.models.ts";
 
-const DEFAULT_AZURE_API_VERSION = "v1";
-const AZURE_TOOL_CALL_PROVIDERS = new Set(["openai", "openai-codex", "opencode", "azure-openai-responses"]);
-
-function parseDeploymentNameMap(value: string | undefined): Map<string, string> {
-	const map = new Map<string, string>();
-	if (!value) return map;
-	for (const entry of value.split(",")) {
-		const trimmed = entry.trim();
-		if (!trimmed) continue;
-		const [modelId, deploymentName] = trimmed.split("=", 2);
-		if (!modelId || !deploymentName) continue;
-		map.set(modelId.trim(), deploymentName.trim());
-	}
-	return map;
-}
-
-function resolveDeploymentName(model: Model<"azure-openai-responses">, options?: AzureOpenAIResponsesOptions): string {
-	if (options?.azureDeploymentName) {
-		return options.azureDeploymentName;
-	}
-	const mappedDeployment = parseDeploymentNameMap(
-		getProviderEnvValue("AZURE_OPENAI_DEPLOYMENT_NAME_MAP", options?.env),
-	).get(model.id);
-	return mappedDeployment || model.id;
-}
-
-function formatAzureOpenAIError(error: unknown): string {
-	if (error instanceof Error) {
-		const status = (error as Error & { status?: unknown }).status;
-		const statusCode = typeof status === "number" ? status : undefined;
-		if (statusCode !== undefined) {
-			return `Azure OpenAI API error (${statusCode}): ${error.message}`;
-		}
-		return error.message;
-	}
-	try {
-		return JSON.stringify(error);
-	} catch {
-		return String(error);
-	}
-}
-
-// Azure OpenAI Responses-specific options
-export interface AzureOpenAIResponsesOptions extends StreamOptions {
-	reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh";
-	reasoningSummary?: "auto" | "detailed" | "concise" | null;
-	azureApiVersion?: string;
-	azureResourceName?: string;
-	azureBaseUrl?: string;
-	azureDeploymentName?: string;
-}
-
-/**
- * Generate function for Azure OpenAI Responses API
- */
-export const streamAzureOpenAIResponses: StreamFunction<"azure-openai-responses", AzureOpenAIResponsesOptions> = (
-	model: Model<"azure-openai-responses">,
-	context: Context,
-	options?: AzureOpenAIResponsesOptions,
-): AssistantMessageEventStream => {
-	const stream = new AssistantMessageEventStream();
-
-	// Start async processing
-	(async () => {
-		const deploymentName = resolveDeploymentName(model, options);
-
-		const output: AssistantMessage = {
-			role: "assistant",
-			content: [],
-			api: "azure-openai-responses" as Api,
-			provider: model.provider,
-			model: model.id,
-			usage: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
-				totalTokens: 0,
-				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
-			},
-			stopReason: "stop",
-			timestamp: Date.now(),
-		};
-
-		try {
-			// Create Azure OpenAI client
-			const apiKey = options?.apiKey;
-			if (!apiKey) {
-				throw new Error(`No API key for provider: ${model.provider}`);
-			}
-			const client = createClient(model, apiKey, options);
-			let params = buildParams(model, context, options, deploymentName);
-			const nextParams = await options?.onPayload?.(params, model);
-			if (nextParams !== undefined) {
-				params = nextParams as ResponseCreateParamsStreaming;
-			}
-			const requestOptions = {
-				...(options?.signal ? { signal: options.signal } : {}),
-				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),
-				maxRetries: options?.maxRetries ?? 0,
-			};
-			const { data: openaiStream, response } = await client.responses.create(params, requestOptions).withResponse();
-			await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
-			stream.push({ type: "start", partial: output });
-
-			await processResponsesStream(openaiStream, output, stream, model);
-
-			if (options?.signal?.aborted) {
-				throw new Error("Request was aborted");
-			}
-
-			if (output.stopReason === "aborted" || output.stopReason === "error") {
-				throw new Error("An unknown error occurred");
-			}
-
-			stream.push({ type: "done", reason: output.stopReason, message: output });
-			stream.end();
-		} catch (error) {
-			for (const block of output.content) {
-				delete (block as { index?: number }).index;
-				// partialJson is only a streaming scratch buffer; never persist it.
-				delete (block as { partialJson?: string }).partialJson;
-			}
-			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = formatAzureOpenAIError(error);
-			stream.push({ type: "error", reason: output.stopReason, error: output });
-			stream.end();
-		}
-	})();
-
-	return stream;
-};
-
-export const streamSimpleAzureOpenAIResponses: StreamFunction<"azure-openai-responses", SimpleStreamOptions> = (
-	model: Model<"azure-openai-responses">,
-	context: Context,
-	options?: SimpleStreamOptions,
-): AssistantMessageEventStream => {
-	const apiKey = options?.apiKey;
-	if (!apiKey) {
-		throw new Error(`No API key for provider: ${model.provider}`);
-	}
-
-	const base = buildBaseOptions(model, options, apiKey);
-	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
-	const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
-
-	return streamAzureOpenAIResponses(model, context, {
-		...base,
-		reasoningEffort,
-	} satisfies AzureOpenAIResponsesOptions);
-};
-
-function normalizeAzureBaseUrl(baseUrl: string): string {
-	const trimmed = baseUrl.trim().replace(/\/+$/, "");
-	let url: URL;
-	try {
-		url = new URL(trimmed);
-	} catch {
-		throw new Error(`Invalid Azure OpenAI base URL: ${baseUrl}`);
-	}
-
-	const isAzureHost =
-		url.hostname.endsWith(".openai.azure.com") || url.hostname.endsWith(".cognitiveservices.azure.com");
-	const normalizedPath = url.pathname.replace(/\/+$/, "");
-
-	// Ensure Azure hosts have /openai/v1 as base path so the AzureOpenAI SDK
-	// can append /deployments/<model>/... and ?api-version=v1 correctly.
-	if (isAzureHost && (normalizedPath === "" || normalizedPath === "/" || normalizedPath === "/openai")) {
-		url.pathname = "/openai/v1";
-		url.search = "";
-	}
-
-	return url.toString().replace(/\/+$/, "");
-}
-
-function buildDefaultBaseUrl(resourceName: string): string {
-	return `https://${resourceName}.openai.azure.com/openai/v1`;
-}
-
-function resolveAzureConfig(
-	model: Model<"azure-openai-responses">,
-	options?: AzureOpenAIResponsesOptions,
-): { baseUrl: string; apiVersion: string } {
-	const apiVersion =
-		options?.azureApiVersion ||
-		getProviderEnvValue("AZURE_OPENAI_API_VERSION", options?.env) ||
-		DEFAULT_AZURE_API_VERSION;
-
-	const baseUrl =
-		options?.azureBaseUrl?.trim() || getProviderEnvValue("AZURE_OPENAI_BASE_URL", options?.env)?.trim() || undefined;
-	const resourceName = options?.azureResourceName || getProviderEnvValue("AZURE_OPENAI_RESOURCE_NAME", options?.env);
-
-	let resolvedBaseUrl = baseUrl;
-
-	if (!resolvedBaseUrl && resourceName) {
-		resolvedBaseUrl = buildDefaultBaseUrl(resourceName);
-	}
-
-	if (!resolvedBaseUrl && model.baseUrl) {
-		resolvedBaseUrl = model.baseUrl;
-	}
-
-	if (!resolvedBaseUrl) {
-		throw new Error(
-			"Azure OpenAI base URL is required. Set AZURE_OPENAI_BASE_URL or AZURE_OPENAI_RESOURCE_NAME, or pass azureBaseUrl, azureResourceName, or model.baseUrl.",
-		);
-	}
-
-	return {
-		baseUrl: normalizeAzureBaseUrl(resolvedBaseUrl),
-		apiVersion,
-	};
-}
-
-function createClient(model: Model<"azure-openai-responses">, apiKey: string, options?: AzureOpenAIResponsesOptions) {
-	const headers = { ...model.headers };
-
-	if (options?.headers) {
-		Object.assign(headers, options.headers);
-	}
-
-	const { baseUrl, apiVersion } = resolveAzureConfig(model, options);
-
-	return new AzureOpenAI({
-		apiKey,
-		apiVersion,
-		dangerouslyAllowBrowser: true,
-		defaultHeaders: headers,
-		baseURL: baseUrl,
+export function azureOpenAIResponsesProvider(): Provider<"azure-openai-responses"> {
+	return createProvider({
+		id: "azure-openai-responses",
+		name: "Azure OpenAI",
+		auth: { apiKey: envApiKeyAuth("Azure OpenAI API key", ["AZURE_OPENAI_API_KEY"]) },
+		models: Object.values(AZURE_OPENAI_RESPONSES_MODELS),
+		api: azureOpenAIResponsesApi(),
 	});
 }
-
-function buildParams(
-	model: Model<"azure-openai-responses">,
-	context: Context,
-	options: AzureOpenAIResponsesOptions | undefined,
-	deploymentName: string,
-) {
-	const messages = convertResponsesMessages(model, context, AZURE_TOOL_CALL_PROVIDERS);
-
-	const params: ResponseCreateParamsStreaming = {
-		model: deploymentName,
-		input: messages,
-		stream: true,
-		prompt_cache_key: clampOpenAIPromptCacheKey(options?.sessionId),
-		store: false,
-	};
-
-	if (options?.maxTokens) {
-		params.max_output_tokens = options?.maxTokens;
-	}
-
-	if (options?.temperature !== undefined) {
-		params.temperature = options?.temperature;
-	}
-
-	if (context.tools && context.tools.length > 0) {
-		params.tools = convertResponsesTools(context.tools);
-	}
-
-	if (model.reasoning) {
-		if (options?.reasoningEffort || options?.reasoningSummary) {
-			const effort = options?.reasoningEffort
-				? (model.thinkingLevelMap?.[options.reasoningEffort] ?? options.reasoningEffort)
-				: "medium";
-			params.reasoning = {
-				effort: effort as NonNullable<typeof params.reasoning>["effort"],
-				summary: options?.reasoningSummary || "auto",
-			};
-			params.include = ["reasoning.encrypted_content"];
-		} else if (model.thinkingLevelMap?.off !== null) {
-			params.reasoning = {
-				effort: (model.thinkingLevelMap?.off ?? "none") as NonNullable<typeof params.reasoning>["effort"],
-			};
-		}
-	}
-
-	return params;
-}

packages/ai/src/providers/cerebras.models.ts

@@ -0,0 +1,41 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const CEREBRAS_MODELS = {
+	"gpt-oss-120b": {
+		id: "gpt-oss-120b",
+		name: "GPT OSS 120B",
+		api: "openai-completions",
+		provider: "cerebras",
+		baseUrl: "https://api.cerebras.ai/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.35,
+			output: 0.75,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 40960,
+	} satisfies Model<"openai-completions">,
+	"zai-glm-4.7": {
+		id: "zai-glm-4.7",
+		name: "Z.AI GLM-4.7",
+		api: "openai-completions",
+		provider: "cerebras",
+		baseUrl: "https://api.cerebras.ai/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 2.25,
+			output: 2.75,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 40960,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/cerebras.ts

@@ -0,0 +1,15 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { CEREBRAS_MODELS } from "./cerebras.models.ts";
+
+export function cerebrasProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "cerebras",
+		name: "Cerebras",
+		baseUrl: "https://api.cerebras.ai/v1",
+		auth: { apiKey: envApiKeyAuth("Cerebras API key", ["CEREBRAS_API_KEY"]) },
+		models: Object.values(CEREBRAS_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/cloudflare-ai-gateway.models.ts

@@ -0,0 +1,656 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const CLOUDFLARE_AI_GATEWAY_MODELS = {
+	"claude-3-5-haiku": {
+		id: "claude-3-5-haiku",
+		name: "Claude Haiku 3.5 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.8,
+			output: 4,
+			cacheRead: 0.08,
+			cacheWrite: 1,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-haiku": {
+		id: "claude-3-haiku",
+		name: "Claude Haiku 3",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.25,
+			cacheRead: 0.03,
+			cacheWrite: 0.3,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-opus": {
+		id: "claude-3-opus",
+		name: "Claude Opus 3",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3-sonnet": {
+		id: "claude-3-sonnet",
+		name: "Claude Sonnet 3",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 0.3,
+		},
+		contextWindow: 200000,
+		maxTokens: 4096,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3.5-haiku": {
+		id: "claude-3.5-haiku",
+		name: "Claude Haiku 3.5 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.8,
+			output: 4,
+			cacheRead: 0.08,
+			cacheWrite: 1,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-3.5-sonnet": {
+		id: "claude-3.5-sonnet",
+		name: "Claude Sonnet 3.5 v2",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 8192,
+	} satisfies Model<"anthropic-messages">,
+	"claude-fable-5": {
+		id: "claude-fable-5",
+		name: "Claude Fable 5",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 50,
+			cacheRead: 1,
+			cacheWrite: 12.5,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-haiku-4-5": {
+		id: "claude-haiku-4-5",
+		name: "Claude Haiku 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1,
+			output: 5,
+			cacheRead: 0.1,
+			cacheWrite: 1.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4": {
+		id: "claude-opus-4",
+		name: "Claude Opus 4 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-1": {
+		id: "claude-opus-4-1",
+		name: "Claude Opus 4.1 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 75,
+			cacheRead: 1.5,
+			cacheWrite: 18.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-5": {
+		id: "claude-opus-4-5",
+		name: "Claude Opus 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-6": {
+		id: "claude-opus-4-6",
+		name: "Claude Opus 4.6 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"max"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-7": {
+		id: "claude-opus-4-7",
+		name: "Claude Opus 4.7",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4-8": {
+		id: "claude-opus-4-8",
+		name: "Claude Opus 4.8",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4": {
+		id: "claude-sonnet-4",
+		name: "Claude Sonnet 4 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-5": {
+		id: "claude-sonnet-4-5",
+		name: "Claude Sonnet 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4-6": {
+		id: "claude-sonnet-4-6",
+		name: "Claude Sonnet 4.6",
+		api: "anthropic-messages",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/anthropic",
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 1000000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"gpt-4": {
+		id: "gpt-4",
+		name: "GPT-4",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 30,
+			output: 60,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 8192,
+		maxTokens: 8192,
+	} satisfies Model<"openai-responses">,
+	"gpt-4-turbo": {
+		id: "gpt-4-turbo",
+		name: "GPT-4 Turbo",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 30,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 4096,
+	} satisfies Model<"openai-responses">,
+	"gpt-4o": {
+		id: "gpt-4o",
+		name: "GPT-4o",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 10,
+			cacheRead: 1.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-responses">,
+	"gpt-4o-mini": {
+		id: "gpt-4o-mini",
+		name: "GPT-4o mini",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.08,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.1": {
+		id: "gpt-5.1",
+		name: "GPT-5.1",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.13,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.1-codex": {
+		id: "gpt-5.1-codex",
+		name: "GPT-5.1 Codex",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.2": {
+		id: "gpt-5.2",
+		name: "GPT-5.2",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.2-codex": {
+		id: "gpt-5.2-codex",
+		name: "GPT-5.2 Codex",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.3-codex": {
+		id: "gpt-5.3-codex",
+		name: "GPT-5.3 Codex",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.4": {
+		id: "gpt-5.4",
+		name: "GPT-5.4",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 15,
+			cacheRead: 0.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.5": {
+		id: "gpt-5.5",
+		name: "GPT-5.5",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 30,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 1050000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"o1": {
+		id: "o1",
+		name: "o1",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 15,
+			output: 60,
+			cacheRead: 7.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"openai-responses">,
+	"o3": {
+		id: "o3",
+		name: "o3",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"openai-responses">,
+	"o3-mini": {
+		id: "o3-mini",
+		name: "o3-mini",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.1,
+			output: 4.4,
+			cacheRead: 0.55,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"openai-responses">,
+	"o3-pro": {
+		id: "o3-pro",
+		name: "o3-pro",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 20,
+			output: 80,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"openai-responses">,
+	"o4-mini": {
+		id: "o4-mini",
+		name: "o4-mini",
+		api: "openai-responses",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/openai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.1,
+			output: 4.4,
+			cacheRead: 0.28,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 100000,
+	} satisfies Model<"openai-responses">,
+	"workers-ai/@cf/moonshotai/kimi-k2.5": {
+		id: "workers-ai/@cf/moonshotai/kimi-k2.5",
+		name: "Kimi K2.5",
+		api: "openai-completions",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/compat",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 3,
+			cacheRead: 0.1,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"openai-completions">,
+	"workers-ai/@cf/moonshotai/kimi-k2.6": {
+		id: "workers-ai/@cf/moonshotai/kimi-k2.6",
+		name: "Kimi K2.6",
+		api: "openai-completions",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/compat",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.16,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"openai-completions">,
+	"workers-ai/@cf/nvidia/nemotron-3-120b-a12b": {
+		id: "workers-ai/@cf/nvidia/nemotron-3-120b-a12b",
+		name: "Nemotron 3 Super 120B",
+		api: "openai-completions",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/compat",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.5,
+			output: 1.5,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"openai-completions">,
+	"workers-ai/@cf/zai-org/glm-4.7-flash": {
+		id: "workers-ai/@cf/zai-org/glm-4.7-flash",
+		name: "GLM-4.7-Flash",
+		api: "openai-completions",
+		provider: "cloudflare-ai-gateway",
+		baseUrl: "https://gateway.ai.cloudflare.com/v1/{CLOUDFLARE_ACCOUNT_ID}/{CLOUDFLARE_GATEWAY_ID}/compat",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.06,
+			output: 0.4,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/cloudflare-ai-gateway.ts

@@ -0,0 +1,22 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { openAIResponsesApi } from "../api/openai-responses.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { CLOUDFLARE_AI_GATEWAY_MODELS } from "./cloudflare-ai-gateway.models.ts";
+
+export function cloudflareAIGatewayProvider(): Provider<
+	"anthropic-messages" | "openai-completions" | "openai-responses"
+> {
+	return createProvider({
+		id: "cloudflare-ai-gateway",
+		name: "Cloudflare AI Gateway",
+		auth: { apiKey: envApiKeyAuth("Cloudflare API key", ["CLOUDFLARE_API_KEY"]) },
+		models: Object.values(CLOUDFLARE_AI_GATEWAY_MODELS),
+		api: {
+			"anthropic-messages": anthropicMessagesApi(),
+			"openai-completions": openAICompletionsApi(),
+			"openai-responses": openAIResponsesApi(),
+		},
+	});
+}

packages/ai/src/providers/cloudflare-workers-ai.models.ts

@@ -0,0 +1,241 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const CLOUDFLARE_WORKERS_AI_MODELS = {
+	"@cf/google/gemma-4-26b-a4b-it": {
+		id: "@cf/google/gemma-4-26b-a4b-it",
+		name: "Gemma 4 26B A4B IT",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"@cf/ibm-granite/granite-4.0-h-micro": {
+		id: "@cf/ibm-granite/granite-4.0-h-micro",
+		name: "Granite 4.0 H Micro",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.017,
+			output: 0.112,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131000,
+		maxTokens: 131000,
+	} satisfies Model<"openai-completions">,
+	"@cf/meta/llama-3.3-70b-instruct-fp8-fast": {
+		id: "@cf/meta/llama-3.3-70b-instruct-fp8-fast",
+		name: "Llama 3.3 70B Instruct fp8 Fast",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.293,
+			output: 2.253,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 24000,
+		maxTokens: 24000,
+	} satisfies Model<"openai-completions">,
+	"@cf/meta/llama-4-scout-17b-16e-instruct": {
+		id: "@cf/meta/llama-4-scout-17b-16e-instruct",
+		name: "Llama 4 Scout 17B 16E Instruct",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.27,
+			output: 0.85,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"@cf/mistralai/mistral-small-3.1-24b-instruct": {
+		id: "@cf/mistralai/mistral-small-3.1-24b-instruct",
+		name: "Mistral Small 3.1 24B Instruct",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.351,
+			output: 0.555,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-completions">,
+	"@cf/moonshotai/kimi-k2.6": {
+		id: "@cf/moonshotai/kimi-k2.6",
+		name: "Kimi K2.6",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.16,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 256000,
+	} satisfies Model<"openai-completions">,
+	"@cf/moonshotai/kimi-k2.7-code": {
+		id: "@cf/moonshotai/kimi-k2.7-code",
+		name: "Kimi K2.7 Code",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.19,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"@cf/nvidia/nemotron-3-120b-a12b": {
+		id: "@cf/nvidia/nemotron-3-120b-a12b",
+		name: "Nemotron 3 Super 120B",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.5,
+			output: 1.5,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"openai-completions">,
+	"@cf/openai/gpt-oss-120b": {
+		id: "@cf/openai/gpt-oss-120b",
+		name: "GPT OSS 120B",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.35,
+			output: 0.75,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"@cf/openai/gpt-oss-20b": {
+		id: "@cf/openai/gpt-oss-20b",
+		name: "GPT OSS 20B",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.2,
+			output: 0.3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"@cf/qwen/qwen3-30b-a3b-fp8": {
+		id: "@cf/qwen/qwen3-30b-a3b-fp8",
+		name: "Qwen3 30B A3b fp8",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.0509,
+			output: 0.335,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 32768,
+		maxTokens: 32768,
+	} satisfies Model<"openai-completions">,
+	"@cf/zai-org/glm-4.7-flash": {
+		id: "@cf/zai-org/glm-4.7-flash",
+		name: "GLM-4.7-Flash",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.0605,
+			output: 0.4,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"@cf/zai-org/glm-5.2": {
+		id: "@cf/zai-org/glm-5.2",
+		name: "Glm 5.2",
+		api: "openai-completions",
+		provider: "cloudflare-workers-ai",
+		baseUrl: "https://api.cloudflare.com/client/v4/accounts/{CLOUDFLARE_ACCOUNT_ID}/ai/v1",
+		compat: {"sendSessionAffinityHeaders":true},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.4,
+			output: 4.4,
+			cacheRead: 0.26,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/cloudflare-workers-ai.ts

@@ -0,0 +1,14 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { CLOUDFLARE_WORKERS_AI_MODELS } from "./cloudflare-workers-ai.models.ts";
+
+export function cloudflareWorkersAIProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "cloudflare-workers-ai",
+		name: "Cloudflare Workers AI",
+		auth: { apiKey: envApiKeyAuth("Cloudflare API key", ["CLOUDFLARE_API_KEY"]) },
+		models: Object.values(CLOUDFLARE_WORKERS_AI_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/deepseek.models.ts

@@ -0,0 +1,45 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const DEEPSEEK_MODELS = {
+	"deepseek-v4-flash": {
+		id: "deepseek-v4-flash",
+		name: "DeepSeek V4 Flash",
+		api: "openai-completions",
+		provider: "deepseek",
+		baseUrl: "https://api.deepseek.com",
+		compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		thinkingLevelMap: {"minimal":null,"low":null,"medium":null,"high":"high","xhigh":"max"},
+		input: ["text"],
+		cost: {
+			input: 0.14,
+			output: 0.28,
+			cacheRead: 0.0028,
+			cacheWrite: 0,
+		},
+		contextWindow: 1000000,
+		maxTokens: 384000,
+	} satisfies Model<"openai-completions">,
+	"deepseek-v4-pro": {
+		id: "deepseek-v4-pro",
+		name: "DeepSeek V4 Pro",
+		api: "openai-completions",
+		provider: "deepseek",
+		baseUrl: "https://api.deepseek.com",
+		compat: {"requiresReasoningContentOnAssistantMessages":true,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		thinkingLevelMap: {"minimal":null,"low":null,"medium":null,"high":"high","xhigh":"max"},
+		input: ["text"],
+		cost: {
+			input: 0.435,
+			output: 0.87,
+			cacheRead: 0.003625,
+			cacheWrite: 0,
+		},
+		contextWindow: 1000000,
+		maxTokens: 384000,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/deepseek.ts

@@ -0,0 +1,15 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { DEEPSEEK_MODELS } from "./deepseek.models.ts";
+
+export function deepseekProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "deepseek",
+		name: "DeepSeek",
+		baseUrl: "https://api.deepseek.com",
+		auth: { apiKey: envApiKeyAuth("DeepSeek API key", ["DEEPSEEK_API_KEY"]) },
+		models: Object.values(DEEPSEEK_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/faux.ts

@@ -1,4 +1,5 @@
 import { registerApiProvider, unregisterApiProviders } from "../api-registry.ts";
+import { createProvider, type Provider } from "../models.ts";
 import type {
 	AssistantMessage,
 	AssistantMessageEventStream,
@@ -125,6 +126,18 @@ export interface FauxProviderRegistration {
 	unregister: () => void;
 }
 
+export interface FauxProviderHandle {
+	provider: Provider;
+	api: string;
+	models: [Model<string>, ...Model<string>[]];
+	getModel(): Model<string>;
+	getModel(modelId: string): Model<string> | undefined;
+	state: { callCount: number };
+	setResponses: (responses: FauxResponseStep[]) => void;
+	appendResponses: (responses: FauxResponseStep[]) => void;
+	getPendingResponseCount: () => number;
+}
+
 function estimateTokens(text: string): number {
 	return Math.ceil(text.length / 4);
 }
@@ -388,10 +401,9 @@ async function streamWithDeltas(
 	stream.end(message);
 }
 
-export function registerFauxProvider(options: RegisterFauxProviderOptions = {}): FauxProviderRegistration {
+function createFauxCore(options: RegisterFauxProviderOptions) {
 	const api = options.api ?? randomId(DEFAULT_API);
 	const provider = options.provider ?? DEFAULT_PROVIDER;
-	const sourceId = randomId("faux-provider");
 	const minTokenSize = Math.max(
 		1,
 		Math.min(options.tokenSize?.min ?? DEFAULT_MIN_TOKEN_SIZE, options.tokenSize?.max ?? DEFAULT_MAX_TOKEN_SIZE),
@@ -467,8 +479,6 @@ export function registerFauxProvider(options: RegisterFauxProviderOptions = {}):
 	const streamSimple: StreamFunction<string, SimpleStreamOptions> = (streamModel, context, streamOptions) =>
 		stream(streamModel, context, streamOptions);
 
-	registerApiProvider({ api, stream, streamSimple }, sourceId);
-
 	function getModel(): Model<string>;
 	function getModel(requestedModelId: string): Model<string> | undefined;
 	function getModel(requestedModelId?: string): Model<string> | undefined {
@@ -480,20 +490,69 @@ export function registerFauxProvider(options: RegisterFauxProviderOptions = {}):
 
 	return {
 		api,
+		provider,
 		models,
+		stream,
+		streamSimple,
 		getModel,
 		state,
-		setResponses(responses) {
+		setResponses(responses: FauxResponseStep[]) {
 			pendingResponses = [...responses];
 		},
-		appendResponses(responses) {
+		appendResponses(responses: FauxResponseStep[]) {
 			pendingResponses.push(...responses);
 		},
 		getPendingResponseCount() {
 			return pendingResponses.length;
 		},
+	};
+}
+
+/** Registers the faux api into the legacy global api-registry. */
+export function registerFauxProvider(options: RegisterFauxProviderOptions = {}): FauxProviderRegistration {
+	const core = createFauxCore(options);
+	const sourceId = randomId("faux-provider");
+	registerApiProvider({ api: core.api, stream: core.stream, streamSimple: core.streamSimple }, sourceId);
+	return {
+		api: core.api,
+		models: core.models,
+		getModel: core.getModel,
+		state: core.state,
+		setResponses: core.setResponses,
+		appendResponses: core.appendResponses,
+		getPendingResponseCount: core.getPendingResponseCount,
 		unregister() {
 			unregisterApiProviders(sourceId);
 		},
 	};
 }
+
+/**
+ * Faux provider for tests built on explicit `Models` collections:
+ *
+ * ```ts
+ * const faux = fauxProvider();
+ * const models = createModels();
+ * models.setProvider(faux.provider);
+ * faux.setResponses([fauxAssistantMessage("hi")]);
+ * ```
+ */
+export function fauxProvider(options: RegisterFauxProviderOptions = {}): FauxProviderHandle {
+	const core = createFauxCore(options);
+	const provider = createProvider({
+		id: core.provider,
+		auth: { apiKey: { name: "Faux", resolve: async () => ({ auth: {} }) } },
+		models: core.models,
+		api: { stream: core.stream, streamSimple: core.streamSimple },
+	});
+	return {
+		provider,
+		api: core.api,
+		models: core.models,
+		getModel: core.getModel,
+		state: core.state,
+		setResponses: core.setResponses,
+		appendResponses: core.appendResponses,
+		getPendingResponseCount: core.getPendingResponseCount,
+	};
+}

packages/ai/src/providers/fireworks.models.ts

@@ -0,0 +1,278 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const FIREWORKS_MODELS = {
+	"accounts/fireworks/models/deepseek-v4-flash": {
+		id: "accounts/fireworks/models/deepseek-v4-flash",
+		name: "DeepSeek V4 Flash",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.14,
+			output: 0.28,
+			cacheRead: 0.028,
+			cacheWrite: 0,
+		},
+		contextWindow: 1000000,
+		maxTokens: 384000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/deepseek-v4-pro": {
+		id: "accounts/fireworks/models/deepseek-v4-pro",
+		name: "DeepSeek V4 Pro",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.74,
+			output: 3.48,
+			cacheRead: 0.145,
+			cacheWrite: 0,
+		},
+		contextWindow: 1000000,
+		maxTokens: 384000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/glm-5p1": {
+		id: "accounts/fireworks/models/glm-5p1",
+		name: "GLM 5.1",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.4,
+			output: 4.4,
+			cacheRead: 0.26,
+			cacheWrite: 0,
+		},
+		contextWindow: 202800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/glm-5p2": {
+		id: "accounts/fireworks/models/glm-5p2",
+		name: "GLM 5.2",
+		api: "openai-completions",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false},
+		reasoning: true,
+		thinkingLevelMap: {"off":"none","minimal":null,"low":"high","medium":"high","xhigh":"max"},
+		input: ["text"],
+		cost: {
+			input: 1.4,
+			output: 4.4,
+			cacheRead: 0.26,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"accounts/fireworks/models/gpt-oss-120b": {
+		id: "accounts/fireworks/models/gpt-oss-120b",
+		name: "GPT OSS 120B",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 32768,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/gpt-oss-20b": {
+		id: "accounts/fireworks/models/gpt-oss-20b",
+		name: "GPT OSS 20B",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.07,
+			output: 0.3,
+			cacheRead: 0.035,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 32768,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/kimi-k2p6": {
+		id: "accounts/fireworks/models/kimi-k2p6",
+		name: "Kimi K2.6",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.16,
+			cacheWrite: 0,
+		},
+		contextWindow: 262000,
+		maxTokens: 262000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/kimi-k2p7-code": {
+		id: "accounts/fireworks/models/kimi-k2p7-code",
+		name: "Kimi K2.7 Code",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.19,
+			cacheWrite: 0,
+		},
+		contextWindow: 262000,
+		maxTokens: 262000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/minimax-m2p7": {
+		id: "accounts/fireworks/models/minimax-m2p7",
+		name: "MiniMax-M2.7",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.06,
+			cacheWrite: 0,
+		},
+		contextWindow: 196608,
+		maxTokens: 196608,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/minimax-m3": {
+		id: "accounts/fireworks/models/minimax-m3",
+		name: "MiniMax-M3",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.06,
+			cacheWrite: 0,
+		},
+		contextWindow: 512000,
+		maxTokens: 512000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/models/qwen3p7-plus": {
+		id: "accounts/fireworks/models/qwen3p7-plus",
+		name: "Qwen 3.7 Plus",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.4,
+			output: 1.6,
+			cacheRead: 0.08,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 65536,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/routers/glm-5p1-fast": {
+		id: "accounts/fireworks/routers/glm-5p1-fast",
+		name: "GLM 5.1 Fast",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 2.8,
+			output: 8.8,
+			cacheRead: 0.52,
+			cacheWrite: 0,
+		},
+		contextWindow: 202800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/routers/kimi-k2p6-fast": {
+		id: "accounts/fireworks/routers/kimi-k2p6-fast",
+		name: "Kimi K2.6 Fast",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.3,
+			cacheWrite: 0,
+		},
+		contextWindow: 262000,
+		maxTokens: 262000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/routers/kimi-k2p6-turbo": {
+		id: "accounts/fireworks/routers/kimi-k2p6-turbo",
+		name: "Kimi K2.6 Turbo",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.3,
+			cacheWrite: 0,
+		},
+		contextWindow: 262000,
+		maxTokens: 262000,
+	} satisfies Model<"anthropic-messages">,
+	"accounts/fireworks/routers/kimi-k2p7-code-fast": {
+		id: "accounts/fireworks/routers/kimi-k2p7-code-fast",
+		name: "Kimi K2.7 Code Fast",
+		api: "anthropic-messages",
+		provider: "fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		compat: {"sendSessionAffinityHeaders":true,"supportsEagerToolInputStreaming":false,"supportsCacheControlOnTools":false,"supportsLongCacheRetention":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.9,
+			output: 8,
+			cacheRead: 0.38,
+			cacheWrite: 0,
+		},
+		contextWindow: 262000,
+		maxTokens: 262000,
+	} satisfies Model<"anthropic-messages">,
+} as const;

packages/ai/src/providers/fireworks.ts

@@ -0,0 +1,19 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { FIREWORKS_MODELS } from "./fireworks.models.ts";
+
+export function fireworksProvider(): Provider<"anthropic-messages" | "openai-completions"> {
+	return createProvider({
+		id: "fireworks",
+		name: "Fireworks",
+		baseUrl: "https://api.fireworks.ai/inference",
+		auth: { apiKey: envApiKeyAuth("Fireworks API key", ["FIREWORKS_API_KEY"]) },
+		models: Object.values(FIREWORKS_MODELS),
+		api: {
+			"anthropic-messages": anthropicMessagesApi(),
+			"openai-completions": openAICompletionsApi(),
+		},
+	});
+}

packages/ai/src/providers/github-copilot.models.ts

@@ -0,0 +1,428 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const GITHUB_COPILOT_MODELS = {
+	"claude-fable-5": {
+		id: "claude-fable-5",
+		name: "Claude Fable 5",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 10,
+			output: 50,
+			cacheRead: 1,
+			cacheWrite: 12.5,
+		},
+		contextWindow: 1000000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-completions">,
+	"claude-haiku-4.5": {
+		id: "claude-haiku-4.5",
+		name: "Claude Haiku 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsEagerToolInputStreaming":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1,
+			output: 5,
+			cacheRead: 0.1,
+			cacheWrite: 1.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4.5": {
+		id: "claude-opus-4.5",
+		name: "Claude Opus 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4.6": {
+		id: "claude-opus-4.6",
+		name: "Claude Opus 4.6",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"max"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 1000000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4.7": {
+		id: "claude-opus-4.7",
+		name: "Claude Opus 4.7",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh","minimal":"low"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-opus-4.8": {
+		id: "claude-opus-4.8",
+		name: "Claude Opus 4.8",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"forceAdaptiveThinking":true,"supportsTemperature":false},
+		reasoning: true,
+		thinkingLevelMap: {"xhigh":"xhigh","minimal":"low"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 25,
+			cacheRead: 0.5,
+			cacheWrite: 6.25,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4": {
+		id: "claude-sonnet-4",
+		name: "Claude Sonnet 4 (latest)",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsEagerToolInputStreaming":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 216000,
+		maxTokens: 16000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4.5": {
+		id: "claude-sonnet-4.5",
+		name: "Claude Sonnet 4.5 (latest)",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsEagerToolInputStreaming":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 200000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"claude-sonnet-4.6": {
+		id: "claude-sonnet-4.6",
+		name: "Claude Sonnet 4.6",
+		api: "anthropic-messages",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"forceAdaptiveThinking":true},
+		reasoning: true,
+		thinkingLevelMap: {"minimal":"low","xhigh":"max"},
+		input: ["text", "image"],
+		cost: {
+			input: 3,
+			output: 15,
+			cacheRead: 0.3,
+			cacheWrite: 3.75,
+		},
+		contextWindow: 1000000,
+		maxTokens: 32000,
+	} satisfies Model<"anthropic-messages">,
+	"gemini-2.5-pro": {
+		id: "gemini-2.5-pro",
+		name: "Gemini 2.5 Pro",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 64000,
+	} satisfies Model<"openai-completions">,
+	"gemini-3-flash-preview": {
+		id: "gemini-3-flash-preview",
+		name: "Gemini 3 Flash Preview",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.5,
+			output: 3,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 64000,
+	} satisfies Model<"openai-completions">,
+	"gemini-3.1-pro-preview": {
+		id: "gemini-3.1-pro-preview",
+		name: "Gemini 3.1 Pro Preview",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"openai-completions">,
+	"gemini-3.5-flash": {
+		id: "gemini-3.5-flash",
+		name: "Gemini 3.5 Flash",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 9,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 64000,
+	} satisfies Model<"openai-completions">,
+	"gpt-4.1": {
+		id: "gpt-4.1",
+		name: "GPT-4.1",
+		api: "openai-completions",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false},
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 8,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"gpt-5-mini": {
+		id: "gpt-5-mini",
+		name: "GPT-5 Mini",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low"},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 2,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 264000,
+		maxTokens: 64000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.2": {
+		id: "gpt-5.2",
+		name: "GPT-5.2",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.2-codex": {
+		id: "gpt-5.2-codex",
+		name: "GPT-5.2 Codex",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.3-codex": {
+		id: "gpt-5.3-codex",
+		name: "GPT-5.3 Codex",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 1.75,
+			output: 14,
+			cacheRead: 0.175,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.4": {
+		id: "gpt-5.4",
+		name: "GPT-5.4",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 2.5,
+			output: 15,
+			cacheRead: 0.25,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.4-mini": {
+		id: "gpt-5.4-mini",
+		name: "GPT-5.4 mini",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 0.75,
+			output: 4.5,
+			cacheRead: 0.075,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.4-nano": {
+		id: "gpt-5.4-nano",
+		name: "GPT-5.4 nano",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 0.2,
+			output: 1.25,
+			cacheRead: 0.02,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+	"gpt-5.5": {
+		id: "gpt-5.5",
+		name: "GPT-5.5",
+		api: "openai-responses",
+		provider: "github-copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		headers: {"User-Agent":"GitHubCopilotChat/0.35.0","Editor-Version":"vscode/1.107.0","Editor-Plugin-Version":"copilot-chat/0.35.0","Copilot-Integration-Id":"vscode-chat"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"low","xhigh":"xhigh"},
+		input: ["text", "image"],
+		cost: {
+			input: 5,
+			output: 30,
+			cacheRead: 0.5,
+			cacheWrite: 0,
+		},
+		contextWindow: 400000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-responses">,
+} as const;

packages/ai/src/providers/github-copilot.ts

@@ -0,0 +1,25 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { openAIResponsesApi } from "../api/openai-responses.lazy.ts";
+import { envApiKeyAuth, lazyOAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { loadGitHubCopilotOAuth } from "../utils/oauth/load.ts";
+import { GITHUB_COPILOT_MODELS } from "./github-copilot.models.ts";
+
+export function githubCopilotProvider(): Provider<"anthropic-messages" | "openai-completions" | "openai-responses"> {
+	return createProvider({
+		id: "github-copilot",
+		name: "GitHub Copilot",
+		baseUrl: "https://api.individual.githubcopilot.com",
+		auth: {
+			apiKey: envApiKeyAuth("GitHub Copilot token", ["COPILOT_GITHUB_TOKEN"]),
+			oauth: lazyOAuth({ name: "GitHub Copilot", load: loadGitHubCopilotOAuth }),
+		},
+		models: Object.values(GITHUB_COPILOT_MODELS),
+		api: {
+			"anthropic-messages": anthropicMessagesApi(),
+			"openai-completions": openAICompletionsApi(),
+			"openai-responses": openAIResponsesApi(),
+		},
+	});
+}

packages/ai/src/providers/google-vertex.models.ts

@@ -0,0 +1,184 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const GOOGLE_VERTEX_MODELS = {
+	"gemini-2.5-flash": {
+		id: "gemini-2.5-flash",
+		name: "Gemini 2.5 Flash",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.3,
+			output: 2.5,
+			cacheRead: 0.03,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-2.5-flash-lite": {
+		id: "gemini-2.5-flash-lite",
+		name: "Gemini 2.5 Flash-Lite",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.4,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-2.5-pro": {
+		id: "gemini-2.5-pro",
+		name: "Gemini 2.5 Pro",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-3-flash-preview": {
+		id: "gemini-3-flash-preview",
+		name: "Gemini 3 Flash Preview",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.5,
+			output: 3,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-3.1-flash-lite": {
+		id: "gemini-3.1-flash-lite",
+		name: "Gemini 3.1 Flash Lite",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.5,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-3.1-pro-preview": {
+		id: "gemini-3.1-pro-preview",
+		name: "Gemini 3.1 Pro Preview",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":"LOW","medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-3.1-pro-preview-customtools": {
+		id: "gemini-3.1-pro-preview-customtools",
+		name: "Gemini 3.1 Pro Preview Custom Tools",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":"LOW","medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-3.5-flash": {
+		id: "gemini-3.5-flash",
+		name: "Gemini 3.5 Flash",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 9,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-flash-latest": {
+		id: "gemini-flash-latest",
+		name: "Gemini Flash Latest",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 9,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+	"gemini-flash-lite-latest": {
+		id: "gemini-flash-lite-latest",
+		name: "Gemini Flash-Lite Latest",
+		api: "google-vertex",
+		provider: "google-vertex",
+		baseUrl: "https://{location}-aiplatform.googleapis.com",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.5,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-vertex">,
+} as const;

packages/ai/src/providers/google-vertex.ts

@@ -1,582 +1,38 @@
-import {
-	type GenerateContentConfig,
-	type GenerateContentParameters,
-	GoogleGenAI,
-	type HttpOptions,
-	ResourceScope,
-	type ThinkingConfig,
-	ThinkingLevel,
-} from "@google/genai";
-import { calculateCost, clampThinkingLevel } from "../models.ts";
-import type {
-	Api,
-	AssistantMessage,
-	Context,
-	Model,
-	ThinkingLevel as PiThinkingLevel,
-	ProviderEnv,
-	SimpleStreamOptions,
-	StreamFunction,
-	StreamOptions,
-	TextContent,
-	ThinkingBudgets,
-	ThinkingContent,
-	ToolCall,
-} from "../types.ts";
-import { AssistantMessageEventStream } from "../utils/event-stream.ts";
-import { getProviderEnvValue } from "../utils/provider-env.ts";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
-import type { GoogleThinkingLevel } from "./google-shared.ts";
-import {
-	convertMessages,
-	convertTools,
-	isThinkingPart,
-	mapStopReason,
-	mapToolChoice,
-	retainThoughtSignature,
-} from "./google-shared.ts";
-import { buildBaseOptions } from "./simple-options.ts";
+import { googleVertexApi } from "../api/google-vertex.lazy.ts";
+import type { ApiKeyAuth } from "../auth/types.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { GOOGLE_VERTEX_MODELS } from "./google-vertex.models.ts";
 
-export interface GoogleVertexOptions extends StreamOptions {
-	toolChoice?: "auto" | "none" | "any";
-	thinking?: {
-		enabled: boolean;
-		budgetTokens?: number; // -1 for dynamic, 0 to disable
-		level?: GoogleThinkingLevel;
-	};
-	project?: string;
-	location?: string;
-}
+const VERTEX_ADC_PATH = "~/.config/gcloud/application_default_credentials.json";
 
-const API_VERSION = "v1";
-const GCP_VERTEX_CREDENTIALS_MARKER = "gcp-vertex-credentials";
+/**
+ * Vertex accepts an explicit API key or Application Default Credentials
+ * (`gcloud auth application-default login`). ADC additionally requires
+ * project and location env vars, which the implementation reads itself.
+ */
+const vertexAuth: ApiKeyAuth = {
+	name: "Google Cloud credentials",
+	resolve: async ({ ctx, credential }) => {
+		const key = credential?.key ?? (await ctx.env("GOOGLE_CLOUD_API_KEY"));
+		if (key) return { auth: { apiKey: key }, source: credential?.key ? "stored credential" : "GOOGLE_CLOUD_API_KEY" };
 
-const THINKING_LEVEL_MAP: Record<GoogleThinkingLevel, ThinkingLevel> = {
-	THINKING_LEVEL_UNSPECIFIED: ThinkingLevel.THINKING_LEVEL_UNSPECIFIED,
-	MINIMAL: ThinkingLevel.MINIMAL,
-	LOW: ThinkingLevel.LOW,
-	MEDIUM: ThinkingLevel.MEDIUM,
-	HIGH: ThinkingLevel.HIGH,
-};
-
-// Counter for generating unique tool call IDs
-let toolCallCounter = 0;
-
-export const streamGoogleVertex: StreamFunction<"google-vertex", GoogleVertexOptions> = (
-	model: Model<"google-vertex">,
-	context: Context,
-	options?: GoogleVertexOptions,
-): AssistantMessageEventStream => {
-	const stream = new AssistantMessageEventStream();
-
-	(async () => {
-		const output: AssistantMessage = {
-			role: "assistant",
-			content: [],
-			api: "google-vertex" as Api,
-			provider: model.provider,
-			model: model.id,
-			usage: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
-				totalTokens: 0,
-				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
-			},
-			stopReason: "stop",
-			timestamp: Date.now(),
-		};
-
-		try {
-			const apiKey = resolveApiKey(options);
-			// Create the client using either a Vertex API key, if provided, or ADC with project and location
-			const client = apiKey
-				? createClientWithApiKey(model, apiKey, options?.headers)
-				: createClient(model, resolveProject(options), resolveLocation(options), options?.headers, options?.env);
-			let params = buildParams(model, context, options);
-			const nextParams = await options?.onPayload?.(params, model);
-			if (nextParams !== undefined) {
-				params = nextParams as GenerateContentParameters;
-			}
-			const googleStream = await client.models.generateContentStream(params);
-
-			stream.push({ type: "start", partial: output });
-			let currentBlock: TextContent | ThinkingContent | null = null;
-			const blocks = output.content;
-			const blockIndex = () => blocks.length - 1;
-			for await (const chunk of googleStream) {
-				// Vertex uses the same @google/genai GenerateContentResponse type as Gemini.
-				// responseId is documented there as an output-only identifier for each response.
-				output.responseId ||= chunk.responseId;
-				const candidate = chunk.candidates?.[0];
-				if (candidate?.content?.parts) {
-					for (const part of candidate.content.parts) {
-						if (part.text !== undefined) {
-							const isThinking = isThinkingPart(part);
-							if (
-								!currentBlock ||
-								(isThinking && currentBlock.type !== "thinking") ||
-								(!isThinking && currentBlock.type !== "text")
-							) {
-								if (currentBlock) {
-									if (currentBlock.type === "text") {
-										stream.push({
-											type: "text_end",
-											contentIndex: blocks.length - 1,
-											content: currentBlock.text,
-											partial: output,
-										});
-									} else {
-										stream.push({
-											type: "thinking_end",
-											contentIndex: blockIndex(),
-											content: currentBlock.thinking,
-											partial: output,
-										});
-									}
-								}
-								if (isThinking) {
-									currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
-									output.content.push(currentBlock);
-									stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
-								} else {
-									currentBlock = { type: "text", text: "" };
-									output.content.push(currentBlock);
-									stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
-								}
-							}
-							if (currentBlock.type === "thinking") {
-								currentBlock.thinking += part.text;
-								currentBlock.thinkingSignature = retainThoughtSignature(
-									currentBlock.thinkingSignature,
-									part.thoughtSignature,
-								);
-								stream.push({
-									type: "thinking_delta",
-									contentIndex: blockIndex(),
-									delta: part.text,
-									partial: output,
-								});
-							} else {
-								currentBlock.text += part.text;
-								currentBlock.textSignature = retainThoughtSignature(
-									currentBlock.textSignature,
-									part.thoughtSignature,
-								);
-								stream.push({
-									type: "text_delta",
-									contentIndex: blockIndex(),
-									delta: part.text,
-									partial: output,
-								});
-							}
-						}
-
-						if (part.functionCall) {
-							if (currentBlock) {
-								if (currentBlock.type === "text") {
-									stream.push({
-										type: "text_end",
-										contentIndex: blockIndex(),
-										content: currentBlock.text,
-										partial: output,
-									});
-								} else {
-									stream.push({
-										type: "thinking_end",
-										contentIndex: blockIndex(),
-										content: currentBlock.thinking,
-										partial: output,
-									});
-								}
-								currentBlock = null;
-							}
-
-							const providedId = part.functionCall.id;
-							const needsNewId =
-								!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
-							const toolCallId = needsNewId
-								? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
-								: providedId;
-
-							const toolCall: ToolCall = {
-								type: "toolCall",
-								id: toolCallId,
-								name: part.functionCall.name || "",
-								arguments: (part.functionCall.args as Record<string, any>) ?? {},
-								...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
-							};
-
-							output.content.push(toolCall);
-							stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
-							stream.push({
-								type: "toolcall_delta",
-								contentIndex: blockIndex(),
-								delta: JSON.stringify(toolCall.arguments),
-								partial: output,
-							});
-							stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
-						}
-					}
-				}
-
-				if (candidate?.finishReason) {
-					output.stopReason = mapStopReason(candidate.finishReason);
-					if (output.content.some((b) => b.type === "toolCall")) {
-						output.stopReason = "toolUse";
-					}
-				}
-
-				if (chunk.usageMetadata) {
-					output.usage = {
-						input:
-							(chunk.usageMetadata.promptTokenCount || 0) - (chunk.usageMetadata.cachedContentTokenCount || 0),
-						output:
-							(chunk.usageMetadata.candidatesTokenCount || 0) + (chunk.usageMetadata.thoughtsTokenCount || 0),
-						cacheRead: chunk.usageMetadata.cachedContentTokenCount || 0,
-						cacheWrite: 0,
-						totalTokens: chunk.usageMetadata.totalTokenCount || 0,
-						cost: {
-							input: 0,
-							output: 0,
-							cacheRead: 0,
-							cacheWrite: 0,
-							total: 0,
-						},
-					};
-					calculateCost(model, output.usage);
-				}
-			}
-
-			if (currentBlock) {
-				if (currentBlock.type === "text") {
-					stream.push({
-						type: "text_end",
-						contentIndex: blockIndex(),
-						content: currentBlock.text,
-						partial: output,
-					});
-				} else {
-					stream.push({
-						type: "thinking_end",
-						contentIndex: blockIndex(),
-						content: currentBlock.thinking,
-						partial: output,
-					});
-				}
-			}
-
-			if (options?.signal?.aborted) {
-				throw new Error("Request was aborted");
-			}
-
-			if (output.stopReason === "aborted" || output.stopReason === "error") {
-				throw new Error("An unknown error occurred");
-			}
-
-			stream.push({ type: "done", reason: output.stopReason, message: output });
-			stream.end();
-		} catch (error) {
-			// Remove internal index property used during streaming
-			for (const block of output.content) {
-				if ("index" in block) {
-					delete (block as { index?: number }).index;
-				}
-			}
-			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
-			stream.push({ type: "error", reason: output.stopReason, error: output });
-			stream.end();
+		const adcPath = await ctx.env("GOOGLE_APPLICATION_CREDENTIALS");
+		const hasCredentials = await ctx.fileExists(adcPath ?? VERTEX_ADC_PATH);
+		const hasProject = Boolean((await ctx.env("GOOGLE_CLOUD_PROJECT")) ?? (await ctx.env("GCLOUD_PROJECT")));
+		const hasLocation = Boolean(await ctx.env("GOOGLE_CLOUD_LOCATION"));
+		if (hasCredentials && hasProject && hasLocation) {
+			return { auth: {}, source: "gcloud application default credentials" };
 		}
-	})();
-
-	return stream;
+		return undefined;
+	},
 };
 
-export const streamSimpleGoogleVertex: StreamFunction<"google-vertex", SimpleStreamOptions> = (
-	model: Model<"google-vertex">,
-	context: Context,
-	options?: SimpleStreamOptions,
-): AssistantMessageEventStream => {
-	const base = buildBaseOptions(model, options, undefined);
-	if (!options?.reasoning) {
-		return streamGoogleVertex(model, context, {
-			...base,
-			thinking: { enabled: false },
-		} satisfies GoogleVertexOptions);
-	}
-
-	const clampedReasoning = clampThinkingLevel(model, options.reasoning);
-	const effort = (clampedReasoning === "off" ? "high" : clampedReasoning) as ClampedThinkingLevel;
-	const geminiModel = model as unknown as Model<"google-generative-ai">;
-
-	if (isGemini3ProModel(geminiModel) || isGemini3FlashModel(geminiModel)) {
-		return streamGoogleVertex(model, context, {
-			...base,
-			thinking: {
-				enabled: true,
-				level: getGemini3ThinkingLevel(effort, geminiModel),
-			},
-		} satisfies GoogleVertexOptions);
-	}
-
-	return streamGoogleVertex(model, context, {
-		...base,
-		thinking: {
-			enabled: true,
-			budgetTokens: getGoogleBudget(geminiModel, effort, options.thinkingBudgets),
-		},
-	} satisfies GoogleVertexOptions);
-};
-
-function createClient(
-	model: Model<"google-vertex">,
-	project: string,
-	location: string,
-	optionsHeaders?: Record<string, string>,
-	env?: ProviderEnv,
-): GoogleGenAI {
-	const googleAuthOptions = buildGoogleAuthOptions(env);
-	return new GoogleGenAI({
-		vertexai: true,
-		project,
-		location,
-		apiVersion: API_VERSION,
-		...(googleAuthOptions ? { googleAuthOptions } : {}),
-		httpOptions: buildHttpOptions(model, optionsHeaders),
+export function googleVertexProvider(): Provider<"google-vertex"> {
+	return createProvider({
+		id: "google-vertex",
+		name: "Google Vertex AI",
+		auth: { apiKey: vertexAuth },
+		models: Object.values(GOOGLE_VERTEX_MODELS),
+		api: googleVertexApi(),
 	});
 }
-
-function createClientWithApiKey(
-	model: Model<"google-vertex">,
-	apiKey: string,
-	optionsHeaders?: Record<string, string>,
-): GoogleGenAI {
-	return new GoogleGenAI({
-		vertexai: true,
-		apiKey,
-		apiVersion: API_VERSION,
-		httpOptions: buildHttpOptions(model, optionsHeaders),
-	});
-}
-
-function buildHttpOptions(
-	model: Model<"google-vertex">,
-	optionsHeaders?: Record<string, string>,
-): HttpOptions | undefined {
-	const httpOptions: HttpOptions = {};
-	const baseUrl = resolveCustomBaseUrl(model.baseUrl);
-	if (baseUrl) {
-		httpOptions.baseUrl = baseUrl;
-		httpOptions.baseUrlResourceScope = ResourceScope.COLLECTION;
-		if (baseUrlIncludesApiVersion(baseUrl)) {
-			httpOptions.apiVersion = "";
-		}
-	}
-
-	if (model.headers || optionsHeaders) {
-		httpOptions.headers = { ...model.headers, ...optionsHeaders };
-	}
-
-	return Object.keys(httpOptions).length > 0 ? httpOptions : undefined;
-}
-
-function resolveCustomBaseUrl(baseUrl: string): string | undefined {
-	const trimmed = baseUrl.trim();
-	if (!trimmed || trimmed.includes("{location}")) {
-		return undefined;
-	}
-	return trimmed;
-}
-
-function baseUrlIncludesApiVersion(baseUrl: string): boolean {
-	try {
-		const url = new URL(baseUrl);
-		return url.pathname.split("/").some((part) => /^v\d+(?:beta\d*)?$/.test(part));
-	} catch {
-		return /(?:^|\/)v\d+(?:beta\d*)?(?:\/|$)/.test(baseUrl);
-	}
-}
-
-function buildGoogleAuthOptions(env?: ProviderEnv): { keyFilename: string } | undefined {
-	const keyFilename = getProviderEnvValue("GOOGLE_APPLICATION_CREDENTIALS", env);
-	return keyFilename ? { keyFilename } : undefined;
-}
-
-function resolveApiKey(options?: GoogleVertexOptions): string | undefined {
-	const apiKey = options?.apiKey?.trim();
-	if (!apiKey || apiKey === GCP_VERTEX_CREDENTIALS_MARKER || isPlaceholderApiKey(apiKey)) {
-		return undefined;
-	}
-	return apiKey;
-}
-
-function isPlaceholderApiKey(apiKey: string): boolean {
-	return /^<[^>]+>$/.test(apiKey);
-}
-
-function resolveProject(options?: GoogleVertexOptions): string {
-	const project =
-		options?.project ||
-		getProviderEnvValue("GOOGLE_CLOUD_PROJECT", options?.env) ||
-		getProviderEnvValue("GCLOUD_PROJECT", options?.env);
-	if (!project) {
-		throw new Error(
-			"Vertex AI requires a project ID. Set GOOGLE_CLOUD_PROJECT/GCLOUD_PROJECT or pass project in options.",
-		);
-	}
-	return project;
-}
-
-function resolveLocation(options?: GoogleVertexOptions): string {
-	const location = options?.location || getProviderEnvValue("GOOGLE_CLOUD_LOCATION", options?.env);
-	if (!location) {
-		throw new Error("Vertex AI requires a location. Set GOOGLE_CLOUD_LOCATION or pass location in options.");
-	}
-	return location;
-}
-
-function buildParams(
-	model: Model<"google-vertex">,
-	context: Context,
-	options: GoogleVertexOptions = {},
-): GenerateContentParameters {
-	const contents = convertMessages(model, context);
-
-	const generationConfig: GenerateContentConfig = {};
-	if (options.temperature !== undefined) {
-		generationConfig.temperature = options.temperature;
-	}
-	if (options.maxTokens !== undefined) {
-		generationConfig.maxOutputTokens = options.maxTokens;
-	}
-
-	const config: GenerateContentConfig = {
-		...(Object.keys(generationConfig).length > 0 && generationConfig),
-		...(context.systemPrompt && { systemInstruction: sanitizeSurrogates(context.systemPrompt) }),
-		...(context.tools && context.tools.length > 0 && { tools: convertTools(context.tools) }),
-	};
-
-	if (context.tools && context.tools.length > 0 && options.toolChoice) {
-		config.toolConfig = {
-			functionCallingConfig: {
-				mode: mapToolChoice(options.toolChoice),
-			},
-		};
-	} else {
-		config.toolConfig = undefined;
-	}
-
-	if (options.thinking?.enabled && model.reasoning) {
-		const thinkingConfig: ThinkingConfig = { includeThoughts: true };
-		if (options.thinking.level !== undefined) {
-			thinkingConfig.thinkingLevel = THINKING_LEVEL_MAP[options.thinking.level];
-		} else if (options.thinking.budgetTokens !== undefined) {
-			thinkingConfig.thinkingBudget = options.thinking.budgetTokens;
-		}
-		config.thinkingConfig = thinkingConfig;
-	} else if (model.reasoning && options.thinking && !options.thinking.enabled) {
-		config.thinkingConfig = getDisabledThinkingConfig(model);
-	}
-
-	if (options.signal) {
-		if (options.signal.aborted) {
-			throw new Error("Request aborted");
-		}
-		config.abortSignal = options.signal;
-	}
-
-	const params: GenerateContentParameters = {
-		model: model.id,
-		contents,
-		config,
-	};
-
-	return params;
-}
-
-type ClampedThinkingLevel = Exclude<PiThinkingLevel, "xhigh">;
-
-function isGemini3ProModel(model: Model<"google-generative-ai">): boolean {
-	return /gemini-3(?:\.\d+)?-pro/.test(model.id.toLowerCase());
-}
-
-function isGemini3FlashModel(model: Model<"google-generative-ai">): boolean {
-	const id = model.id.toLowerCase();
-	return /gemini-3(?:\.\d+)?-flash/.test(id) || id === "gemini-flash-latest" || id === "gemini-flash-lite-latest";
-}
-
-function getDisabledThinkingConfig(model: Model<"google-vertex">): ThinkingConfig {
-	// Google docs: Gemini 3.1 Pro cannot disable thinking, and Gemini 3 Flash / Flash-Lite
-	// do not support full thinking-off either. For Gemini 3 models, use the lowest supported
-	// thinkingLevel without includeThoughts so hidden thinking remains invisible to pi.
-	const geminiModel = model as unknown as Model<"google-generative-ai">;
-	if (isGemini3ProModel(geminiModel)) {
-		return { thinkingLevel: ThinkingLevel.LOW };
-	}
-	if (isGemini3FlashModel(geminiModel)) {
-		return { thinkingLevel: ThinkingLevel.MINIMAL };
-	}
-
-	// Gemini 2.x supports disabling via thinkingBudget = 0.
-	return { thinkingBudget: 0 };
-}
-
-function getGemini3ThinkingLevel(
-	effort: ClampedThinkingLevel,
-	model: Model<"google-generative-ai">,
-): GoogleThinkingLevel {
-	if (isGemini3ProModel(model)) {
-		switch (effort) {
-			case "minimal":
-			case "low":
-				return "LOW";
-			case "medium":
-			case "high":
-				return "HIGH";
-		}
-	}
-	switch (effort) {
-		case "minimal":
-			return "MINIMAL";
-		case "low":
-			return "LOW";
-		case "medium":
-			return "MEDIUM";
-		case "high":
-			return "HIGH";
-	}
-}
-
-function getGoogleBudget(
-	model: Model<"google-generative-ai">,
-	effort: ClampedThinkingLevel,
-	customBudgets?: ThinkingBudgets,
-): number {
-	if (customBudgets?.[effort] !== undefined) {
-		return customBudgets[effort]!;
-	}
-
-	if (model.id.includes("2.5-pro")) {
-		const budgets: Record<ClampedThinkingLevel, number> = {
-			minimal: 128,
-			low: 2048,
-			medium: 8192,
-			high: 32768,
-		};
-		return budgets[effort];
-	}
-
-	if (model.id.includes("2.5-flash")) {
-		const budgets: Record<ClampedThinkingLevel, number> = {
-			minimal: 128,
-			low: 2048,
-			medium: 8192,
-			high: 24576,
-		};
-		return budgets[effort];
-	}
-
-	return -1;
-}

packages/ai/src/providers/google.models.ts

@@ -0,0 +1,290 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const GOOGLE_MODELS = {
+	"gemini-2.0-flash": {
+		id: "gemini-2.0-flash",
+		name: "Gemini 2.0 Flash",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.4,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 8192,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-2.0-flash-lite": {
+		id: "gemini-2.0-flash-lite",
+		name: "Gemini 2.0 Flash-Lite",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.075,
+			output: 0.3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 8192,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-2.5-flash": {
+		id: "gemini-2.5-flash",
+		name: "Gemini 2.5 Flash",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.3,
+			output: 2.5,
+			cacheRead: 0.03,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-2.5-flash-lite": {
+		id: "gemini-2.5-flash-lite",
+		name: "Gemini 2.5 Flash-Lite",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.4,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-2.5-pro": {
+		id: "gemini-2.5-pro",
+		name: "Gemini 2.5 Pro",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.25,
+			output: 10,
+			cacheRead: 0.125,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3-flash-preview": {
+		id: "gemini-3-flash-preview",
+		name: "Gemini 3 Flash Preview",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.5,
+			output: 3,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3-pro-preview": {
+		id: "gemini-3-pro-preview",
+		name: "Gemini 3 Pro Preview",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":"LOW","medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3.1-flash-lite": {
+		id: "gemini-3.1-flash-lite",
+		name: "Gemini 3.1 Flash Lite",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.5,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3.1-flash-lite-preview": {
+		id: "gemini-3.1-flash-lite-preview",
+		name: "Gemini 3.1 Flash Lite Preview",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.5,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3.1-pro-preview": {
+		id: "gemini-3.1-pro-preview",
+		name: "Gemini 3.1 Pro Preview",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":"LOW","medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3.1-pro-preview-customtools": {
+		id: "gemini-3.1-pro-preview-customtools",
+		name: "Gemini 3.1 Pro Preview Custom Tools",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":null,"low":"LOW","medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 12,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-3.5-flash": {
+		id: "gemini-3.5-flash",
+		name: "Gemini 3.5 Flash",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 9,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-flash-latest": {
+		id: "gemini-flash-latest",
+		name: "Gemini Flash Latest",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 9,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemini-flash-lite-latest": {
+		id: "gemini-flash-lite-latest",
+		name: "Gemini Flash-Lite Latest",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.25,
+			output: 1.5,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 65536,
+	} satisfies Model<"google-generative-ai">,
+	"gemma-4-26b-a4b-it": {
+		id: "gemma-4-26b-a4b-it",
+		name: "Gemma 4 26B A4B IT",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"MINIMAL","low":null,"medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"google-generative-ai">,
+	"gemma-4-31b-it": {
+		id: "gemma-4-31b-it",
+		name: "Gemma 4 31B IT",
+		api: "google-generative-ai",
+		provider: "google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		reasoning: true,
+		thinkingLevelMap: {"off":null,"minimal":"MINIMAL","low":null,"medium":null,"high":"HIGH"},
+		input: ["text", "image"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"google-generative-ai">,
+} as const;

packages/ai/src/providers/google.ts

@@ -1,504 +1,15 @@
-import {
-	type GenerateContentConfig,
-	type GenerateContentParameters,
-	GoogleGenAI,
-	type ThinkingConfig,
-} from "@google/genai";
-import { calculateCost, clampThinkingLevel } from "../models.ts";
-import type {
-	Api,
-	AssistantMessage,
-	Context,
-	Model,
-	SimpleStreamOptions,
-	StreamFunction,
-	StreamOptions,
-	TextContent,
-	ThinkingBudgets,
-	ThinkingContent,
-	ThinkingLevel,
-	ToolCall,
-} from "../types.ts";
-import { AssistantMessageEventStream } from "../utils/event-stream.ts";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
-import type { GoogleThinkingLevel } from "./google-shared.ts";
-import {
-	convertMessages,
-	convertTools,
-	isThinkingPart,
-	mapStopReason,
-	mapToolChoice,
-	retainThoughtSignature,
-} from "./google-shared.ts";
-import { buildBaseOptions } from "./simple-options.ts";
+import { googleGenerativeAIApi } from "../api/google-generative-ai.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { GOOGLE_MODELS } from "./google.models.ts";
 
-export interface GoogleOptions extends StreamOptions {
-	toolChoice?: "auto" | "none" | "any";
-	thinking?: {
-		enabled: boolean;
-		budgetTokens?: number; // -1 for dynamic, 0 to disable
-		level?: GoogleThinkingLevel;
-	};
-}
-
-// Counter for generating unique tool call IDs
-let toolCallCounter = 0;
-
-export const streamGoogle: StreamFunction<"google-generative-ai", GoogleOptions> = (
-	model: Model<"google-generative-ai">,
-	context: Context,
-	options?: GoogleOptions,
-): AssistantMessageEventStream => {
-	const stream = new AssistantMessageEventStream();
-
-	(async () => {
-		const output: AssistantMessage = {
-			role: "assistant",
-			content: [],
-			api: "google-generative-ai" as Api,
-			provider: model.provider,
-			model: model.id,
-			usage: {
-				input: 0,
-				output: 0,
-				cacheRead: 0,
-				cacheWrite: 0,
-				totalTokens: 0,
-				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
-			},
-			stopReason: "stop",
-			timestamp: Date.now(),
-		};
-
-		try {
-			const apiKey = options?.apiKey;
-			if (!apiKey) {
-				throw new Error(`No API key for provider: ${model.provider}`);
-			}
-			const client = createClient(model, apiKey, options?.headers);
-			let params = buildParams(model, context, options);
-			const nextParams = await options?.onPayload?.(params, model);
-			if (nextParams !== undefined) {
-				params = nextParams as GenerateContentParameters;
-			}
-			const googleStream = await client.models.generateContentStream(params);
-
-			stream.push({ type: "start", partial: output });
-			let currentBlock: TextContent | ThinkingContent | null = null;
-			const blocks = output.content;
-			const blockIndex = () => blocks.length - 1;
-			for await (const chunk of googleStream) {
-				// @google/genai documents GenerateContentResponse.responseId as an output-only field
-				// used to identify each response. Keep the first non-empty one from the stream.
-				output.responseId ||= chunk.responseId;
-				const candidate = chunk.candidates?.[0];
-				if (candidate?.content?.parts) {
-					for (const part of candidate.content.parts) {
-						if (part.text !== undefined) {
-							const isThinking = isThinkingPart(part);
-							if (
-								!currentBlock ||
-								(isThinking && currentBlock.type !== "thinking") ||
-								(!isThinking && currentBlock.type !== "text")
-							) {
-								if (currentBlock) {
-									if (currentBlock.type === "text") {
-										stream.push({
-											type: "text_end",
-											contentIndex: blocks.length - 1,
-											content: currentBlock.text,
-											partial: output,
-										});
-									} else {
-										stream.push({
-											type: "thinking_end",
-											contentIndex: blockIndex(),
-											content: currentBlock.thinking,
-											partial: output,
-										});
-									}
-								}
-								if (isThinking) {
-									currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
-									output.content.push(currentBlock);
-									stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
-								} else {
-									currentBlock = { type: "text", text: "" };
-									output.content.push(currentBlock);
-									stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
-								}
-							}
-							if (currentBlock.type === "thinking") {
-								currentBlock.thinking += part.text;
-								currentBlock.thinkingSignature = retainThoughtSignature(
-									currentBlock.thinkingSignature,
-									part.thoughtSignature,
-								);
-								stream.push({
-									type: "thinking_delta",
-									contentIndex: blockIndex(),
-									delta: part.text,
-									partial: output,
-								});
-							} else {
-								currentBlock.text += part.text;
-								currentBlock.textSignature = retainThoughtSignature(
-									currentBlock.textSignature,
-									part.thoughtSignature,
-								);
-								stream.push({
-									type: "text_delta",
-									contentIndex: blockIndex(),
-									delta: part.text,
-									partial: output,
-								});
-							}
-						}
-
-						if (part.functionCall) {
-							if (currentBlock) {
-								if (currentBlock.type === "text") {
-									stream.push({
-										type: "text_end",
-										contentIndex: blockIndex(),
-										content: currentBlock.text,
-										partial: output,
-									});
-								} else {
-									stream.push({
-										type: "thinking_end",
-										contentIndex: blockIndex(),
-										content: currentBlock.thinking,
-										partial: output,
-									});
-								}
-								currentBlock = null;
-							}
-
-							// Generate unique ID if not provided or if it's a duplicate
-							const providedId = part.functionCall.id;
-							const needsNewId =
-								!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
-							const toolCallId = needsNewId
-								? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
-								: providedId;
-
-							const toolCall: ToolCall = {
-								type: "toolCall",
-								id: toolCallId,
-								name: part.functionCall.name || "",
-								arguments: (part.functionCall.args as Record<string, any>) ?? {},
-								...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
-							};
-
-							output.content.push(toolCall);
-							stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
-							stream.push({
-								type: "toolcall_delta",
-								contentIndex: blockIndex(),
-								delta: JSON.stringify(toolCall.arguments),
-								partial: output,
-							});
-							stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
-						}
-					}
-				}
-
-				if (candidate?.finishReason) {
-					output.stopReason = mapStopReason(candidate.finishReason);
-					if (output.content.some((b) => b.type === "toolCall")) {
-						output.stopReason = "toolUse";
-					}
-				}
-
-				if (chunk.usageMetadata) {
-					output.usage = {
-						input:
-							(chunk.usageMetadata.promptTokenCount || 0) - (chunk.usageMetadata.cachedContentTokenCount || 0),
-						output:
-							(chunk.usageMetadata.candidatesTokenCount || 0) + (chunk.usageMetadata.thoughtsTokenCount || 0),
-						cacheRead: chunk.usageMetadata.cachedContentTokenCount || 0,
-						cacheWrite: 0,
-						totalTokens: chunk.usageMetadata.totalTokenCount || 0,
-						cost: {
-							input: 0,
-							output: 0,
-							cacheRead: 0,
-							cacheWrite: 0,
-							total: 0,
-						},
-					};
-					calculateCost(model, output.usage);
-				}
-			}
-
-			if (currentBlock) {
-				if (currentBlock.type === "text") {
-					stream.push({
-						type: "text_end",
-						contentIndex: blockIndex(),
-						content: currentBlock.text,
-						partial: output,
-					});
-				} else {
-					stream.push({
-						type: "thinking_end",
-						contentIndex: blockIndex(),
-						content: currentBlock.thinking,
-						partial: output,
-					});
-				}
-			}
-
-			if (options?.signal?.aborted) {
-				throw new Error("Request was aborted");
-			}
-
-			if (output.stopReason === "aborted" || output.stopReason === "error") {
-				throw new Error("An unknown error occurred");
-			}
-
-			stream.push({ type: "done", reason: output.stopReason, message: output });
-			stream.end();
-		} catch (error) {
-			// Remove internal index property used during streaming
-			for (const block of output.content) {
-				if ("index" in block) {
-					delete (block as { index?: number }).index;
-				}
-			}
-			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
-			stream.push({ type: "error", reason: output.stopReason, error: output });
-			stream.end();
-		}
-	})();
-
-	return stream;
-};
-
-export const streamSimpleGoogle: StreamFunction<"google-generative-ai", SimpleStreamOptions> = (
-	model: Model<"google-generative-ai">,
-	context: Context,
-	options?: SimpleStreamOptions,
-): AssistantMessageEventStream => {
-	const apiKey = options?.apiKey;
-	if (!apiKey) {
-		throw new Error(`No API key for provider: ${model.provider}`);
-	}
-
-	const base = buildBaseOptions(model, options, apiKey);
-	if (!options?.reasoning) {
-		return streamGoogle(model, context, { ...base, thinking: { enabled: false } } satisfies GoogleOptions);
-	}
-
-	const clampedReasoning = clampThinkingLevel(model, options.reasoning);
-	const effort = (clampedReasoning === "off" ? "high" : clampedReasoning) as ClampedThinkingLevel;
-	const googleModel = model as Model<"google-generative-ai">;
-
-	if (isGemini3ProModel(googleModel) || isGemini3FlashModel(googleModel) || isGemma4Model(googleModel)) {
-		return streamGoogle(model, context, {
-			...base,
-			thinking: {
-				enabled: true,
-				level: getThinkingLevel(effort, googleModel),
-			},
-		} satisfies GoogleOptions);
-	}
-
-	return streamGoogle(model, context, {
-		...base,
-		thinking: {
-			enabled: true,
-			budgetTokens: getGoogleBudget(googleModel, effort, options.thinkingBudgets),
-		},
-	} satisfies GoogleOptions);
-};
-
-function createClient(
-	model: Model<"google-generative-ai">,
-	apiKey?: string,
-	optionsHeaders?: Record<string, string>,
-): GoogleGenAI {
-	const httpOptions: { baseUrl?: string; apiVersion?: string; headers?: Record<string, string> } = {};
-	if (model.baseUrl) {
-		httpOptions.baseUrl = model.baseUrl;
-		httpOptions.apiVersion = ""; // baseUrl already includes version path, don't append
-	}
-	if (model.headers || optionsHeaders) {
-		httpOptions.headers = { ...model.headers, ...optionsHeaders };
-	}
-
-	return new GoogleGenAI({
-		apiKey,
-		httpOptions: Object.keys(httpOptions).length > 0 ? httpOptions : undefined,
+export function googleProvider(): Provider<"google-generative-ai"> {
+	return createProvider({
+		id: "google",
+		name: "Google",
+		baseUrl: "https://generativelanguage.googleapis.com/v1beta",
+		auth: { apiKey: envApiKeyAuth("Gemini API key", ["GEMINI_API_KEY"]) },
+		models: Object.values(GOOGLE_MODELS),
+		api: googleGenerativeAIApi(),
 	});
 }
-
-function buildParams(
-	model: Model<"google-generative-ai">,
-	context: Context,
-	options: GoogleOptions = {},
-): GenerateContentParameters {
-	const contents = convertMessages(model, context);
-
-	const generationConfig: GenerateContentConfig = {};
-	if (options.temperature !== undefined) {
-		generationConfig.temperature = options.temperature;
-	}
-	if (options.maxTokens !== undefined) {
-		generationConfig.maxOutputTokens = options.maxTokens;
-	}
-
-	const config: GenerateContentConfig = {
-		...(Object.keys(generationConfig).length > 0 && generationConfig),
-		...(context.systemPrompt && { systemInstruction: sanitizeSurrogates(context.systemPrompt) }),
-		...(context.tools && context.tools.length > 0 && { tools: convertTools(context.tools) }),
-	};
-
-	if (context.tools && context.tools.length > 0 && options.toolChoice) {
-		config.toolConfig = {
-			functionCallingConfig: {
-				mode: mapToolChoice(options.toolChoice),
-			},
-		};
-	} else {
-		config.toolConfig = undefined;
-	}
-
-	if (options.thinking?.enabled && model.reasoning) {
-		const thinkingConfig: ThinkingConfig = { includeThoughts: true };
-		if (options.thinking.level !== undefined) {
-			// Cast to any since our GoogleThinkingLevel mirrors Google's ThinkingLevel enum values
-			thinkingConfig.thinkingLevel = options.thinking.level as any;
-		} else if (options.thinking.budgetTokens !== undefined) {
-			thinkingConfig.thinkingBudget = options.thinking.budgetTokens;
-		}
-		config.thinkingConfig = thinkingConfig;
-	} else if (model.reasoning && options.thinking && !options.thinking.enabled) {
-		config.thinkingConfig = getDisabledThinkingConfig(model);
-	}
-
-	if (options.signal) {
-		if (options.signal.aborted) {
-			throw new Error("Request aborted");
-		}
-		config.abortSignal = options.signal;
-	}
-
-	const params: GenerateContentParameters = {
-		model: model.id,
-		contents,
-		config,
-	};
-
-	return params;
-}
-
-type ClampedThinkingLevel = Exclude<ThinkingLevel, "xhigh">;
-
-function isGemma4Model(model: Model<"google-generative-ai">): boolean {
-	return /gemma-?4/.test(model.id.toLowerCase());
-}
-
-function isGemini3ProModel(model: Model<"google-generative-ai">): boolean {
-	return /gemini-3(?:\.\d+)?-pro/.test(model.id.toLowerCase());
-}
-
-function isGemini3FlashModel(model: Model<"google-generative-ai">): boolean {
-	const id = model.id.toLowerCase();
-	return /gemini-3(?:\.\d+)?-flash/.test(id) || id === "gemini-flash-latest" || id === "gemini-flash-lite-latest";
-}
-
-function getDisabledThinkingConfig(model: Model<"google-generative-ai">): ThinkingConfig {
-	// Google docs: Gemini 3.1 Pro cannot disable thinking, and Gemini 3 Flash / Flash-Lite
-	// do not support full thinking-off either. For Gemini 3 models, use the lowest supported
-	// thinkingLevel without includeThoughts so hidden thinking remains invisible to pi.
-	if (isGemini3ProModel(model)) {
-		return { thinkingLevel: "LOW" as any };
-	}
-	if (isGemini3FlashModel(model)) {
-		return { thinkingLevel: "MINIMAL" as any };
-	}
-	if (isGemma4Model(model)) {
-		return { thinkingLevel: "MINIMAL" as any };
-	}
-
-	// Gemini 2.x supports disabling via thinkingBudget = 0.
-	return { thinkingBudget: 0 };
-}
-
-function getThinkingLevel(effort: ClampedThinkingLevel, model: Model<"google-generative-ai">): GoogleThinkingLevel {
-	if (isGemini3ProModel(model)) {
-		switch (effort) {
-			case "minimal":
-			case "low":
-				return "LOW";
-			case "medium":
-			case "high":
-				return "HIGH";
-		}
-	}
-	if (isGemma4Model(model)) {
-		switch (effort) {
-			case "minimal":
-			case "low":
-				return "MINIMAL";
-			case "medium":
-			case "high":
-				return "HIGH";
-		}
-	}
-	switch (effort) {
-		case "minimal":
-			return "MINIMAL";
-		case "low":
-			return "LOW";
-		case "medium":
-			return "MEDIUM";
-		case "high":
-			return "HIGH";
-	}
-}
-
-function getGoogleBudget(
-	model: Model<"google-generative-ai">,
-	effort: ClampedThinkingLevel,
-	customBudgets?: ThinkingBudgets,
-): number {
-	if (customBudgets?.[effort] !== undefined) {
-		return customBudgets[effort]!;
-	}
-
-	if (model.id.includes("2.5-pro")) {
-		const budgets: Record<ClampedThinkingLevel, number> = {
-			minimal: 128,
-			low: 2048,
-			medium: 8192,
-			high: 32768,
-		};
-		return budgets[effort];
-	}
-
-	if (model.id.includes("2.5-flash-lite")) {
-		const budgets: Record<ClampedThinkingLevel, number> = {
-			minimal: 512,
-			low: 2048,
-			medium: 8192,
-			high: 24576,
-		};
-		return budgets[effort];
-	}
-
-	if (model.id.includes("2.5-flash")) {
-		const budgets: Record<ClampedThinkingLevel, number> = {
-			minimal: 128,
-			low: 2048,
-			medium: 8192,
-			high: 24576,
-		};
-		return budgets[effort];
-	}
-
-	return -1;
-}

packages/ai/src/providers/groq.models.ts

@@ -0,0 +1,127 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const GROQ_MODELS = {
+	"llama-3.1-8b-instant": {
+		id: "llama-3.1-8b-instant",
+		name: "Llama 3.1 8B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.05,
+			output: 0.08,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"llama-3.3-70b-versatile": {
+		id: "llama-3.3-70b-versatile",
+		name: "Llama 3.3 70B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.59,
+			output: 0.79,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 32768,
+	} satisfies Model<"openai-completions">,
+	"meta-llama/llama-4-scout-17b-16e-instruct": {
+		id: "meta-llama/llama-4-scout-17b-16e-instruct",
+		name: "Llama 4 Scout 17B 16E",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.11,
+			output: 0.34,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 8192,
+	} satisfies Model<"openai-completions">,
+	"openai/gpt-oss-120b": {
+		id: "openai/gpt-oss-120b",
+		name: "GPT OSS 120B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.075,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"openai/gpt-oss-20b": {
+		id: "openai/gpt-oss-20b",
+		name: "GPT OSS 20B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.075,
+			output: 0.3,
+			cacheRead: 0.0375,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"openai/gpt-oss-safeguard-20b": {
+		id: "openai/gpt-oss-safeguard-20b",
+		name: "Safety GPT OSS 20B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.075,
+			output: 0.3,
+			cacheRead: 0.037,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"qwen/qwen3-32b": {
+		id: "qwen/qwen3-32b",
+		name: "Qwen3-32B",
+		api: "openai-completions",
+		provider: "groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		reasoning: true,
+		thinkingLevelMap: {"minimal":null,"low":null,"medium":null,"high":"default"},
+		input: ["text"],
+		cost: {
+			input: 0.29,
+			output: 0.59,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 40960,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/groq.ts

@@ -0,0 +1,15 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { GROQ_MODELS } from "./groq.models.ts";
+
+export function groqProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "groq",
+		name: "Groq",
+		baseUrl: "https://api.groq.com/openai/v1",
+		auth: { apiKey: envApiKeyAuth("Groq API key", ["GROQ_API_KEY"]) },
+		models: Object.values(GROQ_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/huggingface.models.ts

@@ -0,0 +1,403 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const HUGGINGFACE_MODELS = {
+	"MiniMaxAI/MiniMax-M2.1": {
+		id: "MiniMaxAI/MiniMax-M2.1",
+		name: "MiniMax-M2.1",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"MiniMaxAI/MiniMax-M2.5": {
+		id: "MiniMaxAI/MiniMax-M2.5",
+		name: "MiniMax-M2.5",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.03,
+			cacheWrite: 0,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"MiniMaxAI/MiniMax-M2.7": {
+		id: "MiniMaxAI/MiniMax-M2.7",
+		name: "MiniMax-M2.7",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.06,
+			cacheWrite: 0,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3-235B-A22B-Thinking-2507": {
+		id: "Qwen/Qwen3-235B-A22B-Thinking-2507",
+		name: "Qwen3-235B-A22B-Thinking-2507",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3-Coder-480B-A35B-Instruct": {
+		id: "Qwen/Qwen3-Coder-480B-A35B-Instruct",
+		name: "Qwen3-Coder-480B-A35B-Instruct",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 2,
+			output: 2,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 66536,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3-Coder-Next": {
+		id: "Qwen/Qwen3-Coder-Next",
+		name: "Qwen3-Coder-Next",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.2,
+			output: 1.5,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3-Next-80B-A3B-Instruct": {
+		id: "Qwen/Qwen3-Next-80B-A3B-Instruct",
+		name: "Qwen3-Next-80B-A3B-Instruct",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.25,
+			output: 1,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 66536,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3-Next-80B-A3B-Thinking": {
+		id: "Qwen/Qwen3-Next-80B-A3B-Thinking",
+		name: "Qwen3-Next-80B-A3B-Thinking",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 2,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"Qwen/Qwen3.5-397B-A17B": {
+		id: "Qwen/Qwen3.5-397B-A17B",
+		name: "Qwen3.5-397B-A17B",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 3.6,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"openai-completions">,
+	"XiaomiMiMo/MiMo-V2-Flash": {
+		id: "XiaomiMiMo/MiMo-V2-Flash",
+		name: "MiMo-V2-Flash",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.1,
+			output: 0.3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 4096,
+	} satisfies Model<"openai-completions">,
+	"deepseek-ai/DeepSeek-R1-0528": {
+		id: "deepseek-ai/DeepSeek-R1-0528",
+		name: "DeepSeek-R1-0528",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 3,
+			output: 5,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 163840,
+		maxTokens: 163840,
+	} satisfies Model<"openai-completions">,
+	"deepseek-ai/DeepSeek-V3.2": {
+		id: "deepseek-ai/DeepSeek-V3.2",
+		name: "DeepSeek-V3.2",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.28,
+			output: 0.4,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 163840,
+		maxTokens: 65536,
+	} satisfies Model<"openai-completions">,
+	"deepseek-ai/DeepSeek-V4-Pro": {
+		id: "deepseek-ai/DeepSeek-V4-Pro",
+		name: "DeepSeek V4 Pro",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.435,
+			output: 0.87,
+			cacheRead: 0.003625,
+			cacheWrite: 0,
+		},
+		contextWindow: 1048576,
+		maxTokens: 393216,
+	} satisfies Model<"openai-completions">,
+	"moonshotai/Kimi-K2-Instruct": {
+		id: "moonshotai/Kimi-K2-Instruct",
+		name: "Kimi-K2-Instruct",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 1,
+			output: 3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"moonshotai/Kimi-K2-Instruct-0905": {
+		id: "moonshotai/Kimi-K2-Instruct-0905",
+		name: "Kimi-K2-Instruct-0905",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 1,
+			output: 3,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"moonshotai/Kimi-K2-Thinking": {
+		id: "moonshotai/Kimi-K2-Thinking",
+		name: "Kimi-K2-Thinking",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.5,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"moonshotai/Kimi-K2.5": {
+		id: "moonshotai/Kimi-K2.5",
+		name: "Kimi-K2.5",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 3,
+			cacheRead: 0.1,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"moonshotai/Kimi-K2.6": {
+		id: "moonshotai/Kimi-K2.6",
+		name: "Kimi-K2.6",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.16,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"zai-org/GLM-4.7": {
+		id: "zai-org/GLM-4.7",
+		name: "GLM-4.7",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.2,
+			cacheRead: 0.11,
+			cacheWrite: 0,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"zai-org/GLM-4.7-Flash": {
+		id: "zai-org/GLM-4.7-Flash",
+		name: "GLM-4.7-Flash",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 200000,
+		maxTokens: 128000,
+	} satisfies Model<"openai-completions">,
+	"zai-org/GLM-5": {
+		id: "zai-org/GLM-5",
+		name: "GLM-5",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1,
+			output: 3.2,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 202752,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+	"zai-org/GLM-5.1": {
+		id: "zai-org/GLM-5.1",
+		name: "GLM-5.1",
+		api: "openai-completions",
+		provider: "huggingface",
+		baseUrl: "https://router.huggingface.co/v1",
+		compat: {"supportsDeveloperRole":false},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1,
+			output: 3.2,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 202752,
+		maxTokens: 131072,
+	} satisfies Model<"openai-completions">,
+} as const;

packages/ai/src/providers/huggingface.ts

@@ -0,0 +1,15 @@
+import { openAICompletionsApi } from "../api/openai-completions.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { HUGGINGFACE_MODELS } from "./huggingface.models.ts";
+
+export function huggingfaceProvider(): Provider<"openai-completions"> {
+	return createProvider({
+		id: "huggingface",
+		name: "Hugging Face",
+		baseUrl: "https://router.huggingface.co/v1",
+		auth: { apiKey: envApiKeyAuth("Hugging Face token", ["HF_TOKEN"]) },
+		models: Object.values(HUGGINGFACE_MODELS),
+		api: openAICompletionsApi(),
+	});
+}

packages/ai/src/providers/images/register-builtins.ts

@@ -1,9 +1,9 @@
+import type { generateImages as generateImagesOpenRouterFunction } from "../../api/openrouter-images.ts";
 import { registerImagesApiProvider } from "../../images-api-registry.ts";
 import type { AssistantImages, ImagesContext, ImagesFunction, ImagesModel, ImagesOptions } from "../../types.ts";
-import type { generateImagesOpenRouter as generateImagesOpenRouterFunction } from "./openrouter.ts";
 
 interface OpenRouterImagesProviderModule {
-	generateImagesOpenRouter: typeof generateImagesOpenRouterFunction;
+	generateImages: typeof generateImagesOpenRouterFunction;
 }
 
 let openRouterImagesProviderModulePromise: Promise<OpenRouterImagesProviderModule> | undefined;
@@ -21,7 +21,7 @@ function createLazyLoadErrorImages(model: ImagesModel<"openrouter-images">, erro
 }
 
 function loadOpenRouterImagesProviderModule(): Promise<OpenRouterImagesProviderModule> {
-	openRouterImagesProviderModulePromise ||= import("./openrouter.ts").then(
+	openRouterImagesProviderModulePromise ||= import("../../api/openrouter-images.ts").then(
 		(module) => module as OpenRouterImagesProviderModule,
 	);
 	return openRouterImagesProviderModulePromise;
@@ -34,7 +34,7 @@ export const generateImagesOpenRouter: ImagesFunction<"openrouter-images", Image
 ) => {
 	try {
 		const module = await loadOpenRouterImagesProviderModule();
-		return await module.generateImagesOpenRouter(model, context, options);
+		return await module.generateImages(model, context, options);
 	} catch (error) {
 		return createLazyLoadErrorImages(model, error);
 	}

packages/ai/src/providers/kimi-coding.models.ts

@@ -0,0 +1,61 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const KIMI_CODING_MODELS = {
+	"k2p7": {
+		id: "k2p7",
+		name: "Kimi K2.7 Code",
+		api: "anthropic-messages",
+		provider: "kimi-coding",
+		baseUrl: "https://api.kimi.com/coding",
+		headers: {"User-Agent":"KimiCLI/1.5"},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"anthropic-messages">,
+	"kimi-for-coding": {
+		id: "kimi-for-coding",
+		name: "Kimi For Coding",
+		api: "anthropic-messages",
+		provider: "kimi-coding",
+		baseUrl: "https://api.kimi.com/coding",
+		headers: {"User-Agent":"KimiCLI/1.5"},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"anthropic-messages">,
+	"kimi-k2-thinking": {
+		id: "kimi-k2-thinking",
+		name: "Kimi K2 Thinking",
+		api: "anthropic-messages",
+		provider: "kimi-coding",
+		baseUrl: "https://api.kimi.com/coding",
+		headers: {"User-Agent":"KimiCLI/1.5"},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 32768,
+	} satisfies Model<"anthropic-messages">,
+} as const;

packages/ai/src/providers/kimi-coding.ts

@@ -0,0 +1,15 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { KIMI_CODING_MODELS } from "./kimi-coding.models.ts";
+
+export function kimiCodingProvider(): Provider<"anthropic-messages"> {
+	return createProvider({
+		id: "kimi-coding",
+		name: "Kimi For Coding",
+		baseUrl: "https://api.kimi.com/coding",
+		auth: { apiKey: envApiKeyAuth("Kimi API key", ["KIMI_API_KEY"]) },
+		models: Object.values(KIMI_CODING_MODELS),
+		api: anthropicMessagesApi(),
+	});
+}

packages/ai/src/providers/minimax-cn.models.ts

@@ -0,0 +1,58 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const MINIMAX_CN_MODELS = {
+	"MiniMax-M2.7": {
+		id: "MiniMax-M2.7",
+		name: "MiniMax-M2.7",
+		api: "anthropic-messages",
+		provider: "minimax-cn",
+		baseUrl: "https://api.minimaxi.com/anthropic",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.06,
+			cacheWrite: 0.375,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"MiniMax-M2.7-highspeed": {
+		id: "MiniMax-M2.7-highspeed",
+		name: "MiniMax-M2.7-highspeed",
+		api: "anthropic-messages",
+		provider: "minimax-cn",
+		baseUrl: "https://api.minimaxi.com/anthropic",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.4,
+			cacheRead: 0.06,
+			cacheWrite: 0.375,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"MiniMax-M3": {
+		id: "MiniMax-M3",
+		name: "MiniMax-M3",
+		api: "anthropic-messages",
+		provider: "minimax-cn",
+		baseUrl: "https://api.minimaxi.com/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 2.4,
+			cacheRead: 0.12,
+			cacheWrite: 0,
+		},
+		contextWindow: 512000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+} as const;

packages/ai/src/providers/minimax-cn.ts

@@ -0,0 +1,15 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { MINIMAX_CN_MODELS } from "./minimax-cn.models.ts";
+
+export function minimaxCnProvider(): Provider<"anthropic-messages"> {
+	return createProvider({
+		id: "minimax-cn",
+		name: "MiniMax CN",
+		baseUrl: "https://api.minimaxi.com/anthropic",
+		auth: { apiKey: envApiKeyAuth("MiniMax CN API key", ["MINIMAX_CN_API_KEY"]) },
+		models: Object.values(MINIMAX_CN_MODELS),
+		api: anthropicMessagesApi(),
+	});
+}

packages/ai/src/providers/minimax.models.ts

@@ -0,0 +1,58 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const MINIMAX_MODELS = {
+	"MiniMax-M2.7": {
+		id: "MiniMax-M2.7",
+		name: "MiniMax-M2.7",
+		api: "anthropic-messages",
+		provider: "minimax",
+		baseUrl: "https://api.minimax.io/anthropic",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 1.2,
+			cacheRead: 0.06,
+			cacheWrite: 0.375,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"MiniMax-M2.7-highspeed": {
+		id: "MiniMax-M2.7-highspeed",
+		name: "MiniMax-M2.7-highspeed",
+		api: "anthropic-messages",
+		provider: "minimax",
+		baseUrl: "https://api.minimax.io/anthropic",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.4,
+			cacheRead: 0.06,
+			cacheWrite: 0.375,
+		},
+		contextWindow: 204800,
+		maxTokens: 131072,
+	} satisfies Model<"anthropic-messages">,
+	"MiniMax-M3": {
+		id: "MiniMax-M3",
+		name: "MiniMax-M3",
+		api: "anthropic-messages",
+		provider: "minimax",
+		baseUrl: "https://api.minimax.io/anthropic",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 2.4,
+			cacheRead: 0.12,
+			cacheWrite: 0,
+		},
+		contextWindow: 512000,
+		maxTokens: 128000,
+	} satisfies Model<"anthropic-messages">,
+} as const;

packages/ai/src/providers/minimax.ts

@@ -0,0 +1,15 @@
+import { anthropicMessagesApi } from "../api/anthropic-messages.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { MINIMAX_MODELS } from "./minimax.models.ts";
+
+export function minimaxProvider(): Provider<"anthropic-messages"> {
+	return createProvider({
+		id: "minimax",
+		name: "MiniMax",
+		baseUrl: "https://api.minimax.io/anthropic",
+		auth: { apiKey: envApiKeyAuth("MiniMax API key", ["MINIMAX_API_KEY"]) },
+		models: Object.values(MINIMAX_MODELS),
+		api: anthropicMessagesApi(),
+	});
+}

packages/ai/src/providers/mistral.models.ts

@@ -0,0 +1,517 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const MISTRAL_MODELS = {
+	"codestral-latest": {
+		id: "codestral-latest",
+		name: "Codestral (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.3,
+			output: 0.9,
+			cacheRead: 0.03,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 4096,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-2512": {
+		id: "devstral-2512",
+		name: "Devstral 2",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-latest": {
+		id: "devstral-latest",
+		name: "Devstral 2",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-medium-2507": {
+		id: "devstral-medium-2507",
+		name: "Devstral Medium",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-medium-latest": {
+		id: "devstral-medium-latest",
+		name: "Devstral 2 (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-small-2505": {
+		id: "devstral-small-2505",
+		name: "Devstral Small 2505",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.1,
+			output: 0.3,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"devstral-small-2507": {
+		id: "devstral-small-2507",
+		name: "Devstral Small",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.1,
+			output: 0.3,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"labs-devstral-small-2512": {
+		id: "labs-devstral-small-2512",
+		name: "Devstral Small 2",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"mistral-conversations">,
+	"magistral-medium-latest": {
+		id: "magistral-medium-latest",
+		name: "Magistral Medium (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 2,
+			output: 5,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"mistral-conversations">,
+	"magistral-small": {
+		id: "magistral-small",
+		name: "Magistral Small",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.5,
+			output: 1.5,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"ministral-3b-latest": {
+		id: "ministral-3b-latest",
+		name: "Ministral 3B (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.04,
+			output: 0.04,
+			cacheRead: 0.004,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"ministral-8b-latest": {
+		id: "ministral-8b-latest",
+		name: "Ministral 8B (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.1,
+			output: 0.1,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-large-2411": {
+		id: "mistral-large-2411",
+		name: "Mistral Large 2.1",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 2,
+			output: 6,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 16384,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-large-2512": {
+		id: "mistral-large-2512",
+		name: "Mistral Large 3",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.5,
+			output: 1.5,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-large-latest": {
+		id: "mistral-large-latest",
+		name: "Mistral Large (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.5,
+			output: 1.5,
+			cacheRead: 0.05,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-medium-2505": {
+		id: "mistral-medium-2505",
+		name: "Mistral Medium 3",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 131072,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-medium-2508": {
+		id: "mistral-medium-2508",
+		name: "Mistral Medium 3.1",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-medium-2604": {
+		id: "mistral-medium-2604",
+		name: "Mistral Medium 3.5",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 7.5,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-medium-3.5": {
+		id: "mistral-medium-3.5",
+		name: "Mistral Medium 3.5",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 1.5,
+			output: 7.5,
+			cacheRead: 0,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-medium-latest": {
+		id: "mistral-medium-latest",
+		name: "Mistral Medium (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.4,
+			output: 2,
+			cacheRead: 0.04,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-nemo": {
+		id: "mistral-nemo",
+		name: "Mistral Nemo",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.15,
+			output: 0.15,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-small-2506": {
+		id: "mistral-small-2506",
+		name: "Mistral Small 3.2",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.1,
+			output: 0.3,
+			cacheRead: 0.01,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 16384,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-small-2603": {
+		id: "mistral-small-2603",
+		name: "Mistral Small 4",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"mistral-conversations">,
+	"mistral-small-latest": {
+		id: "mistral-small-latest",
+		name: "Mistral Small (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.15,
+			output: 0.6,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 256000,
+		maxTokens: 256000,
+	} satisfies Model<"mistral-conversations">,
+	"open-mistral-7b": {
+		id: "open-mistral-7b",
+		name: "Mistral 7B",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.25,
+			output: 0.25,
+			cacheRead: 0.025,
+			cacheWrite: 0,
+		},
+		contextWindow: 8000,
+		maxTokens: 8000,
+	} satisfies Model<"mistral-conversations">,
+	"open-mistral-nemo": {
+		id: "open-mistral-nemo",
+		name: "Open Mistral Nemo",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.15,
+			output: 0.15,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"open-mixtral-8x22b": {
+		id: "open-mixtral-8x22b",
+		name: "Mixtral 8x22B",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 2,
+			output: 6,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 64000,
+		maxTokens: 64000,
+	} satisfies Model<"mistral-conversations">,
+	"open-mixtral-8x7b": {
+		id: "open-mixtral-8x7b",
+		name: "Mixtral 8x7B",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.7,
+			output: 0.7,
+			cacheRead: 0.07,
+			cacheWrite: 0,
+		},
+		contextWindow: 32000,
+		maxTokens: 32000,
+	} satisfies Model<"mistral-conversations">,
+	"pixtral-12b": {
+		id: "pixtral-12b",
+		name: "Pixtral 12B",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 0.15,
+			output: 0.15,
+			cacheRead: 0.015,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+	"pixtral-large-latest": {
+		id: "pixtral-large-latest",
+		name: "Pixtral Large (latest)",
+		api: "mistral-conversations",
+		provider: "mistral",
+		baseUrl: "https://api.mistral.ai",
+		reasoning: false,
+		input: ["text", "image"],
+		cost: {
+			input: 2,
+			output: 6,
+			cacheRead: 0.2,
+			cacheWrite: 0,
+		},
+		contextWindow: 128000,
+		maxTokens: 128000,
+	} satisfies Model<"mistral-conversations">,
+} as const;

packages/ai/src/providers/mistral.ts

@@ -1,664 +1,15 @@
-import { Mistral } from "@mistralai/mistralai";
-import type {
-	ChatCompletionStreamRequest,
-	ChatCompletionStreamRequestMessage,
-	CompletionEvent,
-	ContentChunk,
-	FunctionTool,
-} from "@mistralai/mistralai/models/components";
-import { calculateCost, clampThinkingLevel } from "../models.ts";
-import type {
-	AssistantMessage,
-	Context,
-	Message,
-	Model,
-	SimpleStreamOptions,
-	StopReason,
-	StreamFunction,
-	StreamOptions,
-	TextContent,
-	ThinkingContent,
-	Tool,
-	ToolCall,
-} from "../types.ts";
-import { AssistantMessageEventStream } from "../utils/event-stream.ts";
-import { shortHash } from "../utils/hash.ts";
-import { parseStreamingJson } from "../utils/json-parse.ts";
-import { sanitizeSurrogates } from "../utils/sanitize-unicode.ts";
-import { buildBaseOptions } from "./simple-options.ts";
-import { transformMessages } from "./transform-messages.ts";
+import { mistralConversationsApi } from "../api/mistral-conversations.lazy.ts";
+import { envApiKeyAuth } from "../auth/helpers.ts";
+import { createProvider, type Provider } from "../models.ts";
+import { MISTRAL_MODELS } from "./mistral.models.ts";
 
-const MISTRAL_TOOL_CALL_ID_LENGTH = 9;
-const MAX_MISTRAL_ERROR_BODY_CHARS = 4000;
-
-/**
- * Provider-specific options for the Mistral API.
- */
-type MistralReasoningEffort = "none" | "high";
-
-export interface MistralOptions extends StreamOptions {
-	toolChoice?: "auto" | "none" | "any" | "required" | { type: "function"; function: { name: string } };
-	promptMode?: "reasoning";
-	reasoningEffort?: MistralReasoningEffort;
-}
-
-/**
- * Stream responses from Mistral using `chat.stream`.
- */
-export const streamMistral: StreamFunction<"mistral-conversations", MistralOptions> = (
-	model: Model<"mistral-conversations">,
-	context: Context,
-	options?: MistralOptions,
-): AssistantMessageEventStream => {
-	const stream = new AssistantMessageEventStream();
-
-	(async () => {
-		const output = createOutput(model);
-
-		try {
-			const apiKey = options?.apiKey;
-			if (!apiKey) {
-				throw new Error(`No API key for provider: ${model.provider}`);
-			}
-
-			// Intentionally per-request: avoids shared SDK mutable state across concurrent consumers.
-			const mistral = new Mistral({
-				apiKey,
-				serverURL: model.baseUrl,
-			});
-
-			const normalizeMistralToolCallId = createMistralToolCallIdNormalizer();
-			const transformedMessages = transformMessages(context.messages, model, (id) => normalizeMistralToolCallId(id));
-
-			let payload = buildChatPayload(model, context, transformedMessages, options);
-			const nextPayload = await options?.onPayload?.(payload, model);
-			if (nextPayload !== undefined) {
-				payload = nextPayload as ChatCompletionStreamRequest;
-			}
-			const mistralStream = await mistral.chat.stream(payload, buildRequestOptions(model, options));
-			stream.push({ type: "start", partial: output });
-			await consumeChatStream(model, output, stream, mistralStream);
-
-			if (options?.signal?.aborted) {
-				throw new Error("Request was aborted");
-			}
-
-			if (output.stopReason === "aborted" || output.stopReason === "error") {
-				throw new Error("An unknown error occurred");
-			}
-
-			stream.push({ type: "done", reason: output.stopReason, message: output });
-			stream.end();
-		} catch (error) {
-			for (const block of output.content) {
-				// partialArgs is only a streaming scratch buffer; never persist it.
-				delete (block as { partialArgs?: string }).partialArgs;
-			}
-			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = formatMistralError(error);
-			stream.push({ type: "error", reason: output.stopReason, error: output });
-			stream.end();
-		}
-	})();
-
-	return stream;
-};
-
-/**
- * Maps provider-agnostic `SimpleStreamOptions` to Mistral options.
- */
-export const streamSimpleMistral: StreamFunction<"mistral-conversations", SimpleStreamOptions> = (
-	model: Model<"mistral-conversations">,
-	context: Context,
-	options?: SimpleStreamOptions,
-): AssistantMessageEventStream => {
-	const apiKey = options?.apiKey;
-	if (!apiKey) {
-		throw new Error(`No API key for provider: ${model.provider}`);
-	}
-
-	const base = buildBaseOptions(model, options, apiKey);
-	const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
-	const reasoning = clampedReasoning === "off" ? undefined : clampedReasoning;
-	const shouldUseReasoning = model.reasoning && reasoning !== undefined;
-
-	return streamMistral(model, context, {
-		...base,
-		promptMode: shouldUseReasoning && usesPromptModeReasoning(model) ? "reasoning" : undefined,
-		reasoningEffort:
-			shouldUseReasoning && usesReasoningEffort(model) ? mapReasoningEffort(model, reasoning) : undefined,
-	} satisfies MistralOptions);
-};
-
-function createOutput(model: Model<"mistral-conversations">): AssistantMessage {
-	return {
-		role: "assistant",
-		content: [],
-		api: model.api,
-		provider: model.provider,
-		model: model.id,
-		usage: {
-			input: 0,
-			output: 0,
-			cacheRead: 0,
-			cacheWrite: 0,
-			totalTokens: 0,
-			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
-		},
-		stopReason: "stop",
-		timestamp: Date.now(),
-	};
-}
-
-function createMistralToolCallIdNormalizer(): (id: string) => string {
-	const idMap = new Map<string, string>();
-	const reverseMap = new Map<string, string>();
-
-	return (id: string): string => {
-		const existing = idMap.get(id);
-		if (existing) return existing;
-
-		let attempt = 0;
-		while (true) {
-			const candidate = deriveMistralToolCallId(id, attempt);
-			const owner = reverseMap.get(candidate);
-			if (!owner || owner === id) {
-				idMap.set(id, candidate);
-				reverseMap.set(candidate, id);
-				return candidate;
-			}
-			attempt++;
-		}
-	};
-}
-
-function deriveMistralToolCallId(id: string, attempt: number): string {
-	const normalized = id.replace(/[^a-zA-Z0-9]/g, "");
-	if (attempt === 0 && normalized.length === MISTRAL_TOOL_CALL_ID_LENGTH) return normalized;
-	const seedBase = normalized || id;
-	const seed = attempt === 0 ? seedBase : `${seedBase}:${attempt}`;
-	return shortHash(seed)
-		.replace(/[^a-zA-Z0-9]/g, "")
-		.slice(0, MISTRAL_TOOL_CALL_ID_LENGTH);
-}
-
-function formatMistralError(error: unknown): string {
-	if (error instanceof Error) {
-		const sdkError = error as Error & { statusCode?: unknown; body?: unknown };
-		const statusCode = typeof sdkError.statusCode === "number" ? sdkError.statusCode : undefined;
-		const bodyText = typeof sdkError.body === "string" ? sdkError.body.trim() : undefined;
-		if (statusCode !== undefined && bodyText) {
-			return `Mistral API error (${statusCode}): ${truncateErrorText(bodyText, MAX_MISTRAL_ERROR_BODY_CHARS)}`;
-		}
-		if (statusCode !== undefined) return `Mistral API error (${statusCode}): ${error.message}`;
-		return error.message;
-	}
-	return safeJsonStringify(error);
-}
-
-function truncateErrorText(text: string, maxChars: number): string {
-	if (text.length <= maxChars) return text;
-	return `${text.slice(0, maxChars)}... [truncated ${text.length - maxChars} chars]`;
-}
-
-function safeJsonStringify(value: unknown): string {
-	try {
-		const serialized = JSON.stringify(value);
-		return serialized === undefined ? String(value) : serialized;
-	} catch {
-		return String(value);
-	}
-}
-
-function buildRequestOptions(model: Model<"mistral-conversations">, options?: MistralOptions) {
-	const requestOptions: {
-		signal?: AbortSignal;
-		retries: { strategy: "none" };
-		headers?: Record<string, string>;
-	} = {
-		retries: { strategy: "none" },
-	};
-	if (options?.signal) requestOptions.signal = options.signal;
-
-	const headers: Record<string, string> = {};
-	if (model.headers) Object.assign(headers, model.headers);
-	if (options?.headers) Object.assign(headers, options.headers);
-
-	// Mistral infrastructure uses `x-affinity` for KV-cache reuse (prefix caching).
-	// Respect explicit caller-provided header values.
-	if (shouldUsePromptCaching(options) && !headers["x-affinity"]) {
-		headers["x-affinity"] = options.sessionId;
-	}
-
-	if (Object.keys(headers).length > 0) {
-		requestOptions.headers = headers;
-	}
-
-	return requestOptions;
-}
-
-function buildChatPayload(
-	model: Model<"mistral-conversations">,
-	context: Context,
-	messages: Message[],
-	options?: MistralOptions,
-): ChatCompletionStreamRequest {
-	const payload: ChatCompletionStreamRequest = {
-		model: model.id,
-		stream: true,
-		messages: toChatMessages(messages, model.input.includes("image")),
-	};
-
-	if (context.tools?.length) payload.tools = toFunctionTools(context.tools);
-	if (options?.temperature !== undefined) payload.temperature = options.temperature;
-	if (options?.maxTokens !== undefined) payload.maxTokens = options.maxTokens;
-	if (options?.toolChoice) payload.toolChoice = mapToolChoice(options.toolChoice);
-	if (options?.promptMode) payload.promptMode = options.promptMode;
-	if (options?.reasoningEffort) payload.reasoningEffort = options.reasoningEffort;
-	if (shouldUsePromptCaching(options)) payload.promptCacheKey = options.sessionId;
-
-	if (context.systemPrompt) {
-		payload.messages.unshift({
-			role: "system",
-			content: sanitizeSurrogates(context.systemPrompt),
-		});
-	}
-
-	return payload;
-}
-
-function shouldUsePromptCaching(options?: MistralOptions): options is MistralOptions & { sessionId: string } {
-	return options?.cacheRetention !== "none" && !!options?.sessionId;
-}
-
-function getMistralCachedPromptTokens(usage: unknown, promptTokens: number): number {
-	const rawUsage = usage as {
-		promptTokensDetails?: { cachedTokens?: unknown } | null;
-		prompt_tokens_details?: { cached_tokens?: unknown } | null;
-		promptTokenDetails?: { cachedTokens?: unknown } | null;
-		prompt_token_details?: { cached_tokens?: unknown } | null;
-		numCachedTokens?: unknown;
-		num_cached_tokens?: unknown;
-	};
-	const rawCachedTokens =
-		rawUsage.promptTokensDetails?.cachedTokens ??
-		rawUsage.prompt_tokens_details?.cached_tokens ??
-		rawUsage.promptTokenDetails?.cachedTokens ??
-		rawUsage.prompt_token_details?.cached_tokens ??
-		rawUsage.numCachedTokens ??
-		rawUsage.num_cached_tokens ??
-		0;
-	const cachedTokens = typeof rawCachedTokens === "number" && Number.isFinite(rawCachedTokens) ? rawCachedTokens : 0;
-	return Math.min(promptTokens, Math.max(0, cachedTokens));
-}
-
-async function consumeChatStream(
-	model: Model<"mistral-conversations">,
-	output: AssistantMessage,
-	stream: AssistantMessageEventStream,
-	mistralStream: AsyncIterable<CompletionEvent>,
-): Promise<void> {
-	let currentBlock: TextContent | ThinkingContent | null = null;
-	const blocks = output.content;
-	const blockIndex = () => blocks.length - 1;
-	const toolBlocksByKey = new Map<string, number>();
-
-	const finishCurrentBlock = (block?: typeof currentBlock) => {
-		if (!block) return;
-		if (block.type === "text") {
-			stream.push({
-				type: "text_end",
-				contentIndex: blockIndex(),
-				content: block.text,
-				partial: output,
-			});
-			return;
-		}
-		if (block.type === "thinking") {
-			stream.push({
-				type: "thinking_end",
-				contentIndex: blockIndex(),
-				content: block.thinking,
-				partial: output,
-			});
-		}
-	};
-
-	for await (const event of mistralStream) {
-		const chunk = event.data;
-		// Mistral's streamed CompletionChunk carries an id field. Keep the first non-empty one,
-		// mirroring how OpenAI-style streaming exposes a stable response identifier per stream.
-		output.responseId ||= chunk.id;
-
-		if (chunk.usage) {
-			const promptTokens = chunk.usage.promptTokens || 0;
-			const cachedPromptTokens = getMistralCachedPromptTokens(chunk.usage, promptTokens);
-
-			output.usage.input = Math.max(0, promptTokens - cachedPromptTokens);
-			output.usage.output = chunk.usage.completionTokens || 0;
-			output.usage.cacheRead = cachedPromptTokens;
-			output.usage.cacheWrite = 0;
-			output.usage.totalTokens =
-				chunk.usage.totalTokens ||
-				output.usage.input + output.usage.output + output.usage.cacheRead + output.usage.cacheWrite;
-			calculateCost(model, output.usage);
-		}
-
-		const choice = chunk.choices[0];
-		if (!choice) continue;
-
-		if (choice.finishReason) {
-			output.stopReason = mapChatStopReason(choice.finishReason);
-		}
-
-		const delta = choice.delta;
-		if (delta.content !== null && delta.content !== undefined) {
-			const contentItems = typeof delta.content === "string" ? [delta.content] : delta.content;
-			for (const item of contentItems) {
-				if (typeof item === "string") {
-					const textDelta = sanitizeSurrogates(item);
-					if (!currentBlock || currentBlock.type !== "text") {
-						finishCurrentBlock(currentBlock);
-						currentBlock = { type: "text", text: "" };
-						output.content.push(currentBlock);
-						stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
-					}
-					currentBlock.text += textDelta;
-					stream.push({
-						type: "text_delta",
-						contentIndex: blockIndex(),
-						delta: textDelta,
-						partial: output,
-					});
-					continue;
-				}
-
-				if (item.type === "thinking") {
-					const deltaText = item.thinking
-						.map((part) => ("text" in part ? part.text : ""))
-						.filter((text) => text.length > 0)
-						.join("");
-					const thinkingDelta = sanitizeSurrogates(deltaText);
-					if (!thinkingDelta) continue;
-					if (!currentBlock || currentBlock.type !== "thinking") {
-						finishCurrentBlock(currentBlock);
-						currentBlock = { type: "thinking", thinking: "" };
-						output.content.push(currentBlock);
-						stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
-					}
-					currentBlock.thinking += thinkingDelta;
-					stream.push({
-						type: "thinking_delta",
-						contentIndex: blockIndex(),
-						delta: thinkingDelta,
-						partial: output,
-					});
-					continue;
-				}
-
-				if (item.type === "text") {
-					const textDelta = sanitizeSurrogates(item.text);
-					if (!currentBlock || currentBlock.type !== "text") {
-						finishCurrentBlock(currentBlock);
-						currentBlock = { type: "text", text: "" };
-						output.content.push(currentBlock);
-						stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
-					}
-					currentBlock.text += textDelta;
-					stream.push({
-						type: "text_delta",
-						contentIndex: blockIndex(),
-						delta: textDelta,
-						partial: output,
-					});
-				}
-			}
-		}
-
-		const toolCalls = delta.toolCalls || [];
-		for (const toolCall of toolCalls) {
-			if (currentBlock) {
-				finishCurrentBlock(currentBlock);
-				currentBlock = null;
-			}
-			const callId =
-				toolCall.id && toolCall.id !== "null"
-					? toolCall.id
-					: deriveMistralToolCallId(`toolcall:${toolCall.index ?? 0}`, 0);
-			const key = `${callId}:${toolCall.index || 0}`;
-			const existingIndex = toolBlocksByKey.get(key);
-			let block: (ToolCall & { partialArgs?: string }) | undefined;
-
-			if (existingIndex !== undefined) {
-				const existing = output.content[existingIndex];
-				if (existing?.type === "toolCall") {
-					block = existing as ToolCall & { partialArgs?: string };
-				}
-			}
-
-			if (!block) {
-				block = {
-					type: "toolCall",
-					id: callId,
-					name: toolCall.function.name,
-					arguments: {},
-					partialArgs: "",
-				};
-				output.content.push(block);
-				toolBlocksByKey.set(key, output.content.length - 1);
-				stream.push({ type: "toolcall_start", contentIndex: output.content.length - 1, partial: output });
-			}
-
-			const argsDelta =
-				typeof toolCall.function.arguments === "string"
-					? toolCall.function.arguments
-					: JSON.stringify(toolCall.function.arguments || {});
-			block.partialArgs = (block.partialArgs || "") + argsDelta;
-			block.arguments = parseStreamingJson<Record<string, unknown>>(block.partialArgs);
-			stream.push({
-				type: "toolcall_delta",
-				contentIndex: toolBlocksByKey.get(key)!,
-				delta: argsDelta,
-				partial: output,
-			});
-		}
-	}
-
-	finishCurrentBlock(currentBlock);
-	for (const index of toolBlocksByKey.values()) {
-		const block = output.content[index];
-		if (block.type !== "toolCall") continue;
-		const toolBlock = block as ToolCall & { partialArgs?: string };
-		toolBlock.arguments = parseStreamingJson<Record<string, unknown>>(toolBlock.partialArgs);
-		// Finalize in-place and strip the scratch buffer so replay only
-		// carries parsed arguments.
-		delete toolBlock.partialArgs;
-		stream.push({
-			type: "toolcall_end",
-			contentIndex: index,
-			toolCall: toolBlock,
-			partial: output,
-		});
-	}
-}
-
-function toFunctionTools(tools: Tool[]): Array<FunctionTool & { type: "function" }> {
-	return tools.map((tool) => ({
-		type: "function",
-		function: {
-			name: tool.name,
-			description: tool.description,
-			parameters: stripSymbolKeys(tool.parameters) as Record<string, unknown>,
-			strict: false,
-		},
-	}));
-}
-
-function stripSymbolKeys(value: unknown): unknown {
-	if (Array.isArray(value)) {
-		return value.map((item) => stripSymbolKeys(item));
-	}
-
-	if (value && typeof value === "object") {
-		const result: Record<string, unknown> = {};
-		for (const [key, entry] of Object.entries(value)) {
-			result[key] = stripSymbolKeys(entry);
-		}
-		return result;
-	}
-
-	return value;
-}
-
-function toChatMessages(messages: Message[], supportsImages: boolean): ChatCompletionStreamRequestMessage[] {
-	const result: ChatCompletionStreamRequestMessage[] = [];
-
-	for (const msg of messages) {
-		if (msg.role === "user") {
-			if (typeof msg.content === "string") {
-				result.push({ role: "user", content: sanitizeSurrogates(msg.content) });
-				continue;
-			}
-			const hadImages = msg.content.some((item) => item.type === "image");
-			const content: ContentChunk[] = msg.content
-				.filter((item) => item.type === "text" || supportsImages)
-				.map((item) => {
-					if (item.type === "text") return { type: "text", text: sanitizeSurrogates(item.text) };
-					return { type: "image_url", imageUrl: `data:${item.mimeType};base64,${item.data}` };
-				});
-			if (content.length > 0) {
-				result.push({ role: "user", content });
-				continue;
-			}
-			if (hadImages && !supportsImages) {
-				result.push({ role: "user", content: "(image omitted: model does not support images)" });
-			}
-			continue;
-		}
-
-		if (msg.role === "assistant") {
-			const contentParts: ContentChunk[] = [];
-			const toolCalls: Array<{ id: string; type: "function"; function: { name: string; arguments: string } }> = [];
-
-			for (const block of msg.content) {
-				if (block.type === "text") {
-					if (block.text.trim().length > 0) {
-						contentParts.push({ type: "text", text: sanitizeSurrogates(block.text) });
-					}
-					continue;
-				}
-				if (block.type === "thinking") {
-					if (block.thinking.trim().length > 0) {
-						contentParts.push({
-							type: "thinking",
-							thinking: [{ type: "text", text: sanitizeSurrogates(block.thinking) }],
-						});
-					}
-					continue;
-				}
-				toolCalls.push({
-					id: block.id,
-					type: "function",
-					function: { name: block.name, arguments: JSON.stringify(block.arguments || {}) },
-				});
-			}
-
-			const assistantMessage: ChatCompletionStreamRequestMessage = { role: "assistant" };
-			if (contentParts.length > 0) assistantMessage.content = contentParts;
-			if (toolCalls.length > 0) assistantMessage.toolCalls = toolCalls;
-			if (contentParts.length > 0 || toolCalls.length > 0) result.push(assistantMessage);
-			continue;
-		}
-
-		const toolContent: ContentChunk[] = [];
-		const textResult = msg.content
-			.filter((part) => part.type === "text")
-			.map((part) => (part.type === "text" ? sanitizeSurrogates(part.text) : ""))
-			.join("\n");
-		const hasImages = msg.content.some((part) => part.type === "image");
-		const toolText = buildToolResultText(textResult, hasImages, supportsImages, msg.isError);
-		toolContent.push({ type: "text", text: toolText });
-		for (const part of msg.content) {
-			if (!supportsImages) continue;
-			if (part.type !== "image") continue;
-			toolContent.push({
-				type: "image_url",
-				imageUrl: `data:${part.mimeType};base64,${part.data}`,
-			});
-		}
-		result.push({
-			role: "tool",
-			toolCallId: msg.toolCallId,
-			name: msg.toolName,
-			content: toolContent,
-		});
-	}
-
-	return result;
-}
-
-function buildToolResultText(text: string, hasImages: boolean, supportsImages: boolean, isError: boolean): string {
-	const trimmed = text.trim();
-	const errorPrefix = isError ? "[tool error] " : "";
-
-	if (trimmed.length > 0) {
-		const imageSuffix = hasImages && !supportsImages ? "\n[tool image omitted: model does not support images]" : "";
-		return `${errorPrefix}${trimmed}${imageSuffix}`;
-	}
-
-	if (hasImages) {
-		if (supportsImages) {
-			return isError ? "[tool error] (see attached image)" : "(see attached image)";
-		}
-		return isError
-			? "[tool error] (image omitted: model does not support images)"
-			: "(image omitted: model does not support images)";
-	}
-
-	return isError ? "[tool error] (no tool output)" : "(no tool output)";
-}
-
-function usesReasoningEffort(model: Model<"mistral-conversations">): boolean {
-	return model.id === "mistral-small-2603" || model.id === "mistral-small-latest" || model.id === "mistral-medium-3.5";
-}
-
-function usesPromptModeReasoning(model: Model<"mistral-conversations">): boolean {
-	return model.reasoning && !usesReasoningEffort(model);
-}
-
-function mapReasoningEffort(
-	model: Model<"mistral-conversations">,
-	level: Exclude<SimpleStreamOptions["reasoning"], undefined>,
-): MistralReasoningEffort {
-	return (model.thinkingLevelMap?.[level] ?? "high") as MistralReasoningEffort;
-}
-
-function mapToolChoice(
-	choice: MistralOptions["toolChoice"],
-): "auto" | "none" | "any" | "required" | { type: "function"; function: { name: string } } | undefined {
-	if (!choice) return undefined;
-	if (choice === "auto" || choice === "none" || choice === "any" || choice === "required") {
-		return choice as any;
-	}
-	return {
-		type: "function",
-		function: { name: choice.function.name },
-	};
-}
-
-function mapChatStopReason(reason: string | null): StopReason {
-	if (reason === null) return "stop";
-	switch (reason) {
-		case "stop":
-			return "stop";
-		case "length":
-		case "model_length":
-			return "length";
-		case "tool_calls":
-			return "toolUse";
-		case "error":
-			return "error";
-		default:
-			return "stop";
-	}
+export function mistralProvider(): Provider<"mistral-conversations"> {
+	return createProvider({
+		id: "mistral",
+		name: "Mistral",
+		baseUrl: "https://api.mistral.ai",
+		auth: { apiKey: envApiKeyAuth("Mistral API key", ["MISTRAL_API_KEY"]) },
+		models: Object.values(MISTRAL_MODELS),
+		api: mistralConversationsApi(),
+	});
 }

packages/ai/src/providers/moonshotai-cn.models.ts

@@ -0,0 +1,171 @@
+// This file is auto-generated by scripts/generate-models.ts
+// Do not edit manually - run 'npm run generate-models' to update
+
+import type { Model } from "../types.ts";
+
+export const MOONSHOTAI_CN_MODELS = {
+	"kimi-k2-0711-preview": {
+		id: "kimi-k2-0711-preview",
+		name: "Kimi K2 0711",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.5,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 131072,
+		maxTokens: 16384,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2-0905-preview": {
+		id: "kimi-k2-0905-preview",
+		name: "Kimi K2 0905",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.5,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2-thinking": {
+		id: "kimi-k2-thinking",
+		name: "Kimi K2 Thinking",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 0.6,
+			output: 2.5,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2-thinking-turbo": {
+		id: "kimi-k2-thinking-turbo",
+		name: "Kimi K2 Thinking Turbo",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		input: ["text"],
+		cost: {
+			input: 1.15,
+			output: 8,
+			cacheRead: 0.15,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2-turbo-preview": {
+		id: "kimi-k2-turbo-preview",
+		name: "Kimi K2 Turbo",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: false,
+		input: ["text"],
+		cost: {
+			input: 2.4,
+			output: 10,
+			cacheRead: 0.6,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2.5": {
+		id: "kimi-k2.5",
+		name: "Kimi K2.5",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.6,
+			output: 3,
+			cacheRead: 0.1,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2.6": {
+		id: "kimi-k2.6",
+		name: "Kimi K2.6",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.16,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2.7-code": {
+		id: "kimi-k2.7-code",
+		name: "Kimi K2.7 Code",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 0.95,
+			output: 4,
+			cacheRead: 0.19,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+	"kimi-k2.7-code-highspeed": {
+		id: "kimi-k2.7-code-highspeed",
+		name: "Kimi K2.7 Code HighSpeed",
+		api: "openai-completions",
+		provider: "moonshotai-cn",
+		baseUrl: "https://api.moonshot.cn/v1",
+		compat: {"supportsStore":false,"supportsDeveloperRole":false,"supportsReasoningEffort":false,"maxTokensField":"max_tokens","supportsStrictMode":false,"thinkingFormat":"deepseek"},
+		reasoning: true,
+		thinkingLevelMap: {"off":null},
+		input: ["text", "image"],
+		cost: {
+			input: 1.9,
+			output: 8,
+			cacheRead: 0.38,
+			cacheWrite: 0,
+		},
+		contextWindow: 262144,
+		maxTokens: 262144,
+	} satisfies Model<"openai-completions">,
+} as const;