@code-yeongyu/senpi 2026.7.23 → 2026.7.25-2

package/node_modules/@earendil-works/pi-ai/dist/api/openai-codex-responses.js

@@ -15,6 +15,9 @@ import { formatProviderError, normalizeProviderError } from "../utils/error-body
 import { AssistantMessageEventStream } from "../utils/event-stream.js";
 import { headersToRecord } from "../utils/headers.js";
 import { resolveHttpProxyUrlForTarget } from "../utils/node-http-proxy.js";
+import { extractOpenAiCodexAccountId } from "../utils/openai-codex-auth.js";
+import { uuidv7 } from "../utils/uuid.js";
+import { createGrammarToolInputProperties } from "./constrained-sampling.js";
 import { clampOpenAIPromptCacheKey } from "./openai-prompt-cache.js";
 import { convertResponsesMessages, convertResponsesTools, processResponsesStream } from "./openai-responses-shared.js";
 import { applyExtraBody, buildBaseOptions, clampMaxForOpenAI, OPENAI_RESPONSES_RESERVED_BODY_KEYS, } from "./simple-options.js";
@@ -22,7 +25,6 @@ import { applyExtraBody, buildBaseOptions, clampMaxForOpenAI, OPENAI_RESPONSES_R
 // Configuration
 // ============================================================================
 const DEFAULT_CODEX_BASE_URL = "https://chatgpt.com/backend-api";
-const JWT_CLAIM_PATH = "https://api.openai.com/auth";
 const DEFAULT_MAX_RETRIES = 0;
 const BASE_DELAY_MS = 1000;
 const DEFAULT_MAX_RETRY_DELAY_MS = 60_000;
@@ -32,8 +34,9 @@ const DEFAULT_WEBSOCKET_CONNECT_TIMEOUT_MS = 15_000;
 const REQUEST_COMPRESSION_ZSTD_LEVEL = 3;
 const CODEX_TOOL_CALL_PROVIDERS = new Set(["openai", "openai-codex", "opencode"]);
 const WEBSOCKET_MESSAGE_TOO_BIG_CLOSE_CODE = 1009;
-const CODEX_PREVIOUS_RESPONSE_STALE_CODES = new Set(["previous_response_not_found", "codex_previous_response_stale"]);
+const CODEX_PREVIOUS_RESPONSE_STALE_CODES = new Set(["codex_previous_response_stale"]);
 const WEBSOCKET_CONNECTION_LIMIT_REACHED_CODE = "websocket_connection_limit_reached";
+const PREVIOUS_RESPONSE_NOT_FOUND_CODE = "previous_response_not_found";
 const CODEX_RESPONSE_STATUSES = new Set([
     "completed",
     "incomplete",
@@ -79,9 +82,14 @@ function getRetryAfterDelayMs(headers) {
     }
     return undefined;
 }
-function capRetryDelayMs(delayMs, options) {
+class RetryDelayExceededError extends Error {
+}
+function validateRetryDelayMs(delayMs, options) {
     const maxRetryDelayMs = options?.maxRetryDelayMs ?? DEFAULT_MAX_RETRY_DELAY_MS;
-    return maxRetryDelayMs > 0 ? Math.min(delayMs, maxRetryDelayMs) : delayMs;
+    if (maxRetryDelayMs > 0 && delayMs > maxRetryDelayMs) {
+        throw new RetryDelayExceededError(`Server requested ${Math.ceil(delayMs / 1000)}s retry delay (max: ${Math.ceil(maxRetryDelayMs / 1000)}s)`);
+    }
+    return delayMs;
 }
 function sleep(ms, signal) {
     return new Promise((resolve, reject) => {
@@ -162,32 +170,40 @@ export const stream = (model, context, options) => {
                 throw new Error(`No API key for provider: ${model.provider}`);
             }
             const accountId = extractAccountId(apiKey);
-            let body = buildRequestBody(model, context, options);
+            const grammarToolInputProperties = createGrammarToolInputProperties(context.tools, model.compat?.supportsOpenAIGrammarTools ?? false);
+            const cacheSessionId = options?.cacheRetention === "none" ? undefined : options?.sessionId;
+            const codexSessionId = clampOpenAIPromptCacheKey(cacheSessionId);
+            let body = buildRequestBody(model, context, options, codexSessionId, grammarToolInputProperties);
             const nextBody = await options?.onPayload?.(body, model);
             if (nextBody !== undefined) {
                 body = nextBody;
             }
-            const codexSessionId = clampOpenAIPromptCacheKey(options?.sessionId);
-            const websocketRequestId = codexSessionId || createCodexRequestId();
+            const websocketRequestId = codexSessionId || uuidv7();
             const sseHeaders = buildSSEHeaders(model.headers, options?.headers, accountId, apiKey, codexSessionId);
             const websocketHeaders = buildWebSocketHeaders(model.headers, options?.headers, accountId, apiKey, websocketRequestId);
             const bodyJson = JSON.stringify(body);
             const httpTimeoutMs = normalizeTimeoutMs(options?.timeoutMs);
             const websocketConnectTimeoutMs = normalizeTimeoutMs(options?.websocketConnectTimeoutMs);
             const transport = options?.transport || "auto";
-            const websocketDisabledForSession = transport !== "sse" && isWebSocketSseFallbackActive(options?.sessionId);
+            let startEmitted = false;
+            const websocketDisabledForSession = transport !== "sse" && isWebSocketSseFallbackActive(cacheSessionId);
             if (websocketDisabledForSession) {
-                recordWebSocketSseFallback(options?.sessionId);
+                recordWebSocketSseFallback(cacheSessionId);
             }
             if (transport !== "sse" && !websocketDisabledForSession) {
                 let websocketStarted = false;
                 let retriedWebSocketConnectionLimit = false;
+                let retriedMissingWebSocketContinuation = false;
                 while (true) {
                     websocketStarted = false;
                     try {
                         await processWebSocketStream(resolveCodexWebSocketUrl(model.baseUrl), body, websocketHeaders, output, stream, model, () => {
                             websocketStarted = true;
-                        }, httpTimeoutMs, websocketConnectTimeoutMs, options);
+                            if (!startEmitted) {
+                                startEmitted = true;
+                                stream.push({ type: "start", partial: output });
+                            }
+                        }, httpTimeoutMs, websocketConnectTimeoutMs, cacheSessionId, grammarToolInputProperties, options);
                         if (options?.signal?.aborted) {
                             throw new Error("Request was aborted");
                         }
@@ -202,6 +218,11 @@ export const stream = (model, context, options) => {
                     catch (error) {
                         const aborted = options?.signal?.aborted;
                         const connectionLimitBeforeStart = !websocketStarted && isWebSocketConnectionLimitReachedError(error);
+                        const previousResponseNotFound = isPreviousResponseNotFoundError(error);
+                        if (!aborted && previousResponseNotFound && !retriedMissingWebSocketContinuation) {
+                            retriedMissingWebSocketContinuation = true;
+                            continue;
+                        }
                         if (!aborted && connectionLimitBeforeStart && !retriedWebSocketConnectionLimit) {
                             retriedWebSocketConnectionLimit = true;
                             continue;
@@ -216,11 +237,11 @@ export const stream = (model, context, options) => {
                             phase: websocketStarted ? "after_message_stream_start" : "before_message_stream_start",
                             requestBytes: new TextEncoder().encode(bodyJson).byteLength,
                         }));
-                        recordWebSocketFailure(options?.sessionId, error);
+                        recordWebSocketFailure(cacheSessionId, error);
                         if (websocketStarted) {
                             throw error;
                         }
-                        recordWebSocketSseFallback(options?.sessionId);
+                        recordWebSocketSseFallback(cacheSessionId);
                         break;
                     }
                 }
@@ -270,9 +291,7 @@ export const stream = (model, context, options) => {
                         const retryAfterDelayMs = getRetryAfterDelayMs(response.headers);
                         const delayMs = retryAfterDelayMs === undefined
                             ? BASE_DELAY_MS * 2 ** attempt
-                            : response.status === 429
-                                ? capRetryDelayMs(retryAfterDelayMs, options)
-                                : retryAfterDelayMs;
+                            : validateRetryDelayMs(retryAfterDelayMs, options);
                         await sleep(delayMs, options?.signal);
                         continue;
                     }
@@ -292,7 +311,9 @@ export const stream = (model, context, options) => {
                     }
                     lastError = error instanceof Error ? error : new Error(String(error));
                     // Network errors are retryable
-                    if (attempt < maxRetries && !lastError.message.includes("usage limit")) {
+                    if (attempt < maxRetries &&
+                        !(lastError instanceof RetryDelayExceededError) &&
+                        !lastError.message.includes("usage limit")) {
                         const delayMs = BASE_DELAY_MS * 2 ** attempt;
                         await sleep(delayMs, options?.signal);
                         continue;
@@ -306,8 +327,11 @@ export const stream = (model, context, options) => {
             if (!response.body) {
                 throw new Error("No response body");
             }
-            stream.push({ type: "start", partial: output });
-            await processStream(response, output, stream, model, options);
+            if (!startEmitted) {
+                startEmitted = true;
+                stream.push({ type: "start", partial: output });
+            }
+            await processStream(response, output, stream, model, grammarToolInputProperties, options);
             if (options?.signal?.aborted) {
                 throw new Error("Request was aborted");
             }
@@ -316,8 +340,9 @@ export const stream = (model, context, options) => {
         }
         catch (error) {
             for (const block of output.content) {
-                // partialJson is only a streaming scratch buffer; never persist it.
+                // Streaming scratch buffers are only used during parsing; never persist them.
                 delete block.partialJson;
+                delete block.customInput;
             }
             output.stopReason = options?.signal?.aborted ? "aborted" : "error";
             output.errorMessage = formatProviderError(normalizeProviderError(error));
@@ -347,14 +372,29 @@ export const streamSimple = (model, context, options) => {
 // ============================================================================
 // Request Building
 // ============================================================================
-function buildRequestBody(model, context, options) {
-    const reasoningRequested = options?.reasoningEffort !== undefined && options.reasoningEffort !== "none";
+function buildRequestBody(model, context, options, cacheSessionId, grammarToolInputProperties = createGrammarToolInputProperties(context.tools, model.compat?.supportsOpenAIGrammarTools ?? false)) {
+    const requestedReasoningEffort = options?.reasoningEffort;
+    const mappedReasoningEffort = requestedReasoningEffort === undefined
+        ? undefined
+        : requestedReasoningEffort === "none"
+            ? model.thinkingLevelMap?.off
+            : model.thinkingLevelMap?.[requestedReasoningEffort];
+    const reasoningEffort = mappedReasoningEffort === undefined ? requestedReasoningEffort : mappedReasoningEffort;
+    const reasoningRequested = requestedReasoningEffort !== undefined && requestedReasoningEffort !== "none" && reasoningEffort !== null;
+    const supportsStrictMode = model.compat?.supportsStrictMode ?? true;
+    const supportsOpenAIGrammarTools = model.compat?.supportsOpenAIGrammarTools ?? false;
     const toolPlacement = splitDeferredTools(context, model.compat?.supportsToolSearch ?? false);
     const messages = convertResponsesMessages(model, context, CODEX_TOOL_CALL_PROVIDERS, {
         includeSystemPrompt: false,
         preserveThinking: reasoningRequested,
         preserveTextSignatures: true,
+        grammarToolInputProperties,
         deferredTools: toolPlacement.deferred,
+        toolOptions: {
+            strict: null,
+            supportsStrictMode,
+            supportsOpenAIGrammarTools,
+        },
     });
     const body = {
         model: model.id,
@@ -364,7 +404,7 @@ function buildRequestBody(model, context, options) {
         input: messages,
         text: { verbosity: options?.textVerbosity || "low" },
         include: ["reasoning.encrypted_content"],
-        prompt_cache_key: clampOpenAIPromptCacheKey(options?.sessionId),
+        prompt_cache_key: cacheSessionId,
         tool_choice: options?.toolChoice ?? "auto",
         parallel_tool_calls: true,
     };
@@ -375,18 +415,23 @@ function buildRequestBody(model, context, options) {
         body.service_tier = options.serviceTier;
     }
     if (toolPlacement.immediate.length > 0) {
-        body.tools = convertResponsesTools(toolPlacement.immediate, { strict: null });
+        body.tools = convertResponsesTools(toolPlacement.immediate, {
+            strict: null,
+            supportsStrictMode,
+            supportsOpenAIGrammarTools,
+        });
     }
-    if (options?.reasoningEffort !== undefined) {
-        const effort = options.reasoningEffort === "none"
-            ? (model.thinkingLevelMap?.off ?? "none")
-            : (model.thinkingLevelMap?.[options.reasoningEffort] ?? options.reasoningEffort);
-        if (effort !== null) {
-            body.reasoning = {
-                effort,
-                summary: options.reasoningSummary ?? "auto",
-            };
-        }
+    if (reasoningEffort !== undefined && reasoningEffort !== null) {
+        body.reasoning = {
+            effort: reasoningEffort,
+            summary: options?.reasoningSummary ?? "auto",
+        };
+    }
+    else if (reasoningEffort === undefined && model.reasoning && model.thinkingLevelMap?.off !== null) {
+        body.reasoning = {
+            effort: model.thinkingLevelMap?.off ?? "none",
+            summary: options?.reasoningSummary ?? "auto",
+        };
     }
     applyExtraBody(body, options?.extraBody, OPENAI_RESPONSES_RESERVED_BODY_KEYS);
     return body;
@@ -437,9 +482,10 @@ function resolveCodexWebSocketUrl(baseUrl) {
 // ============================================================================
 // Response Processing
 // ============================================================================
-async function processStream(response, output, stream, model, options) {
+async function processStream(response, output, stream, model, grammarToolInputProperties, options) {
     await processResponsesStream(mapCodexEvents(parseSSE(response, options?.signal)), output, stream, model, {
         serviceTier: options?.serviceTier,
+        grammarToolInputProperties,
         resolveServiceTier: resolveCodexServiceTier,
         applyServiceTierPricing: (usage, serviceTier) => applyServiceTierPricing(usage, serviceTier, model),
     });
@@ -488,6 +534,9 @@ function isCodexPreviousResponseStale(error) {
         typeof error.code === "string" &&
         CODEX_PREVIOUS_RESPONSE_STALE_CODES.has(error.code));
 }
+function isPreviousResponseNotFoundError(error) {
+    return error instanceof CodexApiError && error.code === PREVIOUS_RESPONSE_NOT_FOUND_CODE;
+}
 function isWebSocketConnectionLimitReachedError(error) {
     return error instanceof CodexApiError && error.code === WEBSOCKET_CONNECTION_LIMIT_REACHED_CODE;
 }
@@ -1125,34 +1174,34 @@ function recordWebSocketRequestStats(stats, requestBody, reused, useCachedContex
         stats.lastPreviousResponseId = undefined;
     }
 }
-async function* startWebSocketOutputOnFirstEvent(events, output, stream, onStart) {
+async function* startWebSocketOutputOnFirstEvent(events, onStart) {
     let started = false;
     for await (const event of events) {
         if (!started) {
             started = true;
             onStart();
-            stream.push({ type: "start", partial: output });
         }
         yield event;
     }
 }
-async function processWebSocketStream(url, body, headers, output, stream, model, onStart, idleTimeoutMs, websocketConnectTimeoutMs, options) {
-    const { socket, entry, reused, release } = await acquireWebSocket(url, headers, options?.sessionId, options?.signal, websocketConnectTimeoutMs, options?.env);
+async function processWebSocketStream(url, body, headers, output, stream, model, onStart, idleTimeoutMs, websocketConnectTimeoutMs, cacheSessionId, grammarToolInputProperties, options) {
+    const { socket, entry, reused, release } = await acquireWebSocket(url, headers, cacheSessionId, options?.signal, websocketConnectTimeoutMs, options?.env);
     let keepConnection = true;
     const useCachedContext = options?.transport === "websocket-cached" || options?.transport === "auto";
     // ChatGPT Codex Responses rejects `store: true` ("Store must be set to false").
     // WebSocket continuation still works via connection-scoped previous_response_id state.
     const fullBody = body;
     let requestBody = useCachedContext && entry ? buildCachedWebSocketRequestBody(entry, fullBody) : fullBody;
-    const stats = options?.sessionId ? getOrCreateWebSocketDebugStats(options.sessionId) : undefined;
+    const stats = cacheSessionId ? getOrCreateWebSocketDebugStats(cacheSessionId) : undefined;
     let recoveredStalePreviousResponse = false;
     try {
         while (true) {
             recordWebSocketRequestStats(stats, requestBody, reused && !recoveredStalePreviousResponse, useCachedContext);
             try {
                 socket.send(JSON.stringify({ type: "response.create", ...requestBody }));
-                await processResponsesStream(startWebSocketOutputOnFirstEvent(mapCodexEvents(parseWebSocket(socket, options?.signal, idleTimeoutMs)), output, stream, onStart), output, stream, model, {
+                await processResponsesStream(startWebSocketOutputOnFirstEvent(mapCodexEvents(parseWebSocket(socket, options?.signal, idleTimeoutMs)), onStart), output, stream, model, {
                     serviceTier: options?.serviceTier,
+                    grammarToolInputProperties,
                     resolveServiceTier: resolveCodexServiceTier,
                     applyServiceTierPricing: (usage, serviceTier) => applyServiceTierPricing(usage, serviceTier, model),
                 });
@@ -1178,7 +1227,8 @@ async function processWebSocketStream(url, body, headers, output, stream, model,
         else if (useCachedContext && entry && output.responseId) {
             const responseItems = convertResponsesMessages(model, { messages: [output] }, CODEX_TOOL_CALL_PROVIDERS, {
                 includeSystemPrompt: false,
-            }).filter((item) => item.type !== "function_call_output");
+                grammarToolInputProperties,
+            }).filter((item) => item.type !== "function_call_output" && item.type !== "custom_tool_call_output");
             entry.continuation = {
                 lastRequestBody: fullBody,
                 lastResponseId: output.responseId,
@@ -1227,23 +1277,7 @@ async function parseErrorResponse(response) {
 // Auth & Headers
 // ============================================================================
 function extractAccountId(token) {
-    try {
-        const parts = token.split(".");
-        if (parts.length !== 3)
-            return undefined;
-        const payload = JSON.parse(atob(parts[1]));
-        const accountId = payload?.[JWT_CLAIM_PATH]?.chatgpt_account_id;
-        return typeof accountId === "string" && accountId.length > 0 ? accountId : undefined;
-    }
-    catch {
-        return undefined;
-    }
-}
-function createCodexRequestId() {
-    if (typeof globalThis.crypto?.randomUUID === "function") {
-        return globalThis.crypto.randomUUID();
-    }
-    return `codex_${Date.now()}_${Math.random().toString(36).slice(2, 10)}`;
+    return extractOpenAiCodexAccountId(token);
 }
 function buildBaseCodexHeaders(initHeaders, additionalHeaders, accountId, token) {
     const headers = new Headers(initHeaders);