322 lines
10 KiB
TypeScript
322 lines
10 KiB
TypeScript
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 { formatProviderError, normalizeProviderError } from "../utils/error-body.ts";
|
|
import { AssistantMessageEventStream } from "../utils/event-stream.ts";
|
|
import { headersToRecord } from "../utils/headers.ts";
|
|
import { getProviderEnvValue } from "../utils/provider-env.ts";
|
|
import { retryProviderRequest } from "../utils/provider-retry.ts";
|
|
import { createGrammarToolInputProperties } from "./constrained-sampling.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"]);
|
|
// OpenAI Responses rejects max_output_tokens below 16: https://github.com/earendil-works/pi/issues/6265
|
|
const OPENAI_RESPONSES_MIN_OUTPUT_TOKENS = 16;
|
|
|
|
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 {
|
|
return formatProviderError(normalizeProviderError(error), "Azure OpenAI API error");
|
|
}
|
|
|
|
// Azure OpenAI Responses-specific options
|
|
export interface AzureOpenAIResponsesOptions extends StreamOptions {
|
|
reasoningEffort?: "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
|
|
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);
|
|
const grammarToolInputProperties = createGrammarToolInputProperties(
|
|
context.tools,
|
|
model.compat?.supportsOpenAIGrammarTools ?? false,
|
|
);
|
|
let params = buildParams(model, context, options, deploymentName, grammarToolInputProperties);
|
|
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: 0,
|
|
};
|
|
const { data: openaiStream, response } = await retryProviderRequest(
|
|
() => client.responses.create(params, requestOptions).withResponse(),
|
|
{
|
|
maxRetries: options?.maxRetries,
|
|
maxRetryDelayMs: options?.maxRetryDelayMs,
|
|
signal: options?.signal,
|
|
},
|
|
);
|
|
await options?.onResponse?.({ status: response.status, headers: headersToRecord(response.headers) }, model);
|
|
stream.push({ type: "start", partial: output });
|
|
|
|
await processResponsesStream(openaiStream, output, stream, model, { grammarToolInputProperties });
|
|
|
|
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;
|
|
// Streaming scratch buffers are only used during parsing; never persist them.
|
|
delete (block as { partialJson?: string }).partialJson;
|
|
delete (block as { customInput?: unknown }).customInput;
|
|
}
|
|
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: ${model.provider}`);
|
|
}
|
|
|
|
const base = buildBaseOptions(model, context, options, apiKey);
|
|
const clampedReasoning = options?.reasoning ? clampThinkingLevel(model, options.reasoning) : undefined;
|
|
const reasoningEffort = clampedReasoning === "off" ? undefined : clampedReasoning;
|
|
|
|
return stream(model, context, {
|
|
...base,
|
|
reasoningEffort,
|
|
} satisfies AzureOpenAIResponsesOptions);
|
|
};
|
|
|
|
function normalizeAzureBaseUrl(baseUrl: string): string {
|
|
const trimmed = baseUrl.trim().replace(/\/+$/, "");
|
|
let url: URL;
|
|
try {
|
|
url = new URL(trimmed);
|
|
} catch {
|
|
throw new Error(`Invalid Azure OpenAI base URL: ${baseUrl}`);
|
|
}
|
|
|
|
const isAzureHost =
|
|
url.hostname.endsWith(".openai.azure.com") ||
|
|
url.hostname.endsWith(".cognitiveservices.azure.com") ||
|
|
url.hostname.endsWith(".ai.azure.com");
|
|
const normalizedPath = url.pathname.replace(/\/+$/, "");
|
|
|
|
// 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" ||
|
|
normalizedPath === "/openai/v1/responses")
|
|
) {
|
|
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,
|
|
grammarToolInputProperties: ReadonlyMap<string, string> = createGrammarToolInputProperties(
|
|
context.tools,
|
|
model.compat?.supportsOpenAIGrammarTools ?? false,
|
|
),
|
|
) {
|
|
const messages = convertResponsesMessages(model, context, AZURE_TOOL_CALL_PROVIDERS, {
|
|
grammarToolInputProperties,
|
|
});
|
|
|
|
const params: ResponseCreateParamsStreaming = {
|
|
model: deploymentName,
|
|
input: messages,
|
|
stream: true,
|
|
prompt_cache_key: clampOpenAIPromptCacheKey(options?.sessionId),
|
|
store: false,
|
|
};
|
|
|
|
if (options?.maxTokens) {
|
|
params.max_output_tokens = Math.max(options.maxTokens, OPENAI_RESPONSES_MIN_OUTPUT_TOKENS);
|
|
}
|
|
|
|
if (options?.temperature !== undefined) {
|
|
params.temperature = options?.temperature;
|
|
}
|
|
|
|
if (context.tools && context.tools.length > 0) {
|
|
params.tools = convertResponsesTools(context.tools, {
|
|
supportsStrictMode: model.compat?.supportsStrictMode ?? true,
|
|
supportsOpenAIGrammarTools: model.compat?.supportsOpenAIGrammarTools ?? false,
|
|
});
|
|
}
|
|
|
|
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;
|
|
}
|