Move provider auth and OAuth flows onto pi-ai Models, compose models.json and extension overlays through ModelRuntime, and retain ModelRegistry as an extension compatibility facade.
Provider.getModels() is sync-only (last-known list; must not throw) with
an optional refreshModels() where dynamic providers fetch. The
sync-or-async union invited latent sync assumptions that would detonate
on the first dynamic provider; async-only reads would force sync
consumer surfaces (extension find/getAll) through Promises. Sync reads
plus an explicit refresh verb keeps the contract single and the
staleness visible.
Models.getModels()/getModel() are sync best-effort reads;
Models.refresh(provider?) rejects with ModelsError(model_source) for a
single provider and is concurrent best-effort across all providers.
createProvider() takes a models array plus an optional refreshModels
fetcher (stored on success, in-flight calls deduped, list unchanged on
rejection). forceRefresh options are gone.
Also finishes the in-progress AuthStorage fallbackResolver removal
(drops the now-unused includeFallback option from getApiKey).
The root barrel is now core-only and side-effect free: types,
createModels/createProvider, auth substrate, lazyStream/lazyApi, faux,
utils. Generated catalogs, api-registry, env-api-keys, images, global
stream functions, and per-API lazy wrappers leave the root.
New @earendil-works/pi-ai/compat preserves the old surface verbatim as
a strict superset of the root: api-dispatch stream/complete with env
key injection, the builtin registration side effect (skip-if-present so
it cannot clobber earlier overrides), deprecated getModel/getModels/
getProviders aliases of the new getBuiltin* reads in providers/all,
lazy api wrappers + setBedrockProviderModule, and image generation.
Compat dies with the coding-agent ModelManager migration.
Packaging: exports map gains ./compat, ./providers/*, ./api/*;
sideEffects array lists only the effectful modules.
Old-global imports across agent/coding-agent/examples and pi-ai tests
switch to /compat (path-only; compat is a superset). The coding-agent
extension loader resolves the pi-ai ROOT specifier to compat, so
existing user extensions using the old global API keep working at
runtime until compat is removed. vitest configs alias /compat to src;
browser smoke imports old globals from /compat.
* feat(coding-agent): allow comments and trailing commas in models.json
Run user-supplied models.json through a small `stripJsonComments` helper
before JSON.parse so users can annotate their config and leave trailing
commas without breaking the loader.
Co-Authored-By: julien-agent <Agents+cyolo@huggingface.co>
* fix(coding-agent): strip comments before trailing commas in models.json
The single-pass regex couldn't see a trailing comma when a `//` comment sat
between the comma and its closer. Split into two passes: strip comments
first, then strip trailing commas on the cleaned input.
Co-Authored-By: julien-agent <Agents+cyolo@huggingface.co>
---------
Co-authored-by: julien-agent <Agents+cyolo@huggingface.co>
Previously, registerProvider() overwrote the stored config for a provider,
so a models registration followed by a baseUrl/headers-only override lost
the models. After refresh(), only the override-only config was replayed,
causing extension-provided models to disappear.
Added upsertRegisteredProvider() that merges defined fields into the
existing stored config instead of replacing it. Fields absent from the
incoming config are preserved from the stored config.
Also adds regression tests for dynamic provider override persistence
across refresh.
* fix(typebox): migrate to v1 with extension compat
Replace AJV-based validation with TypeBox-native validation, keep legacy extension imports working (including @sinclair/typebox/compiler), and restore coercion for serialized/plain JSON schemas.
This change closes#3112.
* fix(typebox): use canonical imports and harden coercion
Switch first-party code to canonical typebox imports while retaining legacy extension aliases in the loader.
Remove obsolete runtime codegen guards, expand serialized JSON-schema coercion coverage, and update related tests and fixtures.
Fixes#3112.
---------
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
* fix(coding-agent): surface models.json load errors on stderr in --list-models
When models.json has validation errors, --list-models silently
discarded custom models and overrides without any user-visible
feedback. Now prints the error to stderr via chalk.yellow warning
before listing available models.
* fix(coding-agent): allow custom models for built-in providers in models.json
Built-in providers (openrouter, anthropic, etc.) already have api and
baseUrl on every model, and auth comes from env vars / auth storage.
Relax validation so custom models under built-in providers don't need
redundant baseUrl, apiKey, or api fields. Inherit them from the first
built-in model for that provider.
fixes#2921
---------
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
The OpenRouterRouting interface only had only and order, but the runtime
passes the entire object as-is to the OpenRouter API's provider field.
This means all other fields worked at runtime but had no TypeScript
type-checking, no autocomplete, and no config validation.
This adds all fields from the OpenRouter provider selection API
(https://openrouter.ai/docs/guides/routing/provider-selection):
allow_fallbacks, require_parameters, data_collection, zdr,
enforce_distillable_text, ignore, quantizations, sort, max_price,
preferred_min_throughput, preferred_max_latency.
Also fixes the models.md documentation which had an invalid fallbacks
field in the example (not part of the OpenRouter API), replaces it with
a comprehensive example showing all available fields, and clarifies that
openRouterRouting is sent as-is in the provider field.
API keys in auth.json now support the same resolution as models.json:
- Shell command: "\!command" executes and uses stdout (cached)
- Environment variable: uses the value of the named variable
- Literal value: used directly
Extracted shared resolveConfigValue() to new resolve-config-value.ts module.
Model definitions now only require 'id'. All other fields have sensible
defaults for local models (Ollama, LM Studio, etc.):
- name: defaults to id
- reasoning: defaults to false
- input: defaults to ["text"]
- cost: defaults to {input: 0, output: 0, cacheRead: 0, cacheWrite: 0}
- contextWindow: defaults to 128000
- maxTokens: defaults to 16384
Existing configs that specify all fields continue to work unchanged.
Fixes#1146
Mario Zechner