n9router 0.4.43 → 0.4.45

package/app/open-sse/executors/antigravity.js

@@ -3,9 +3,10 @@ import { BaseExecutor } from "./base.js";
 import { PROVIDERS } from "../config/providers.js";
 import { OAUTH_ENDPOINTS, ANTIGRAVITY_HEADERS, INTERNAL_REQUEST_HEADER, AG_DEFAULT_TOOLS, AG_TOOL_SUFFIX } from "../config/appConstants.js";
 import { HTTP_STATUS, DEFAULT_AG_503_RETRY_COUNT } from "../config/runtimeConfig.js";
-import { deriveSessionId } from "../utils/sessionManager.js";
+import { resolveSessionId } from "../utils/sessionManager.js";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
-import { cleanJSONSchemaForAntigravity } from "../translator/helpers/geminiHelper.js";
+import { cleanJSONSchemaForAntigravity } from "../translator/formats/gemini.js";
+import { DEFAULT_THINKING_AG_SIGNATURE } from "../config/defaultThinkingSignature.js";
 
 // Sanitize function name: Gemini requires [a-zA-Z_][a-zA-Z0-9_.:\-]{0,63}
 function sanitizeFunctionName(name) {
@@ -16,31 +17,112 @@ function sanitizeFunctionName(name) {
 }
 
 const MAX_RETRY_AFTER_MS = 10000;
+const ANTIGRAVITY_TRANSIENT_RETRY_MAX_MS = 15000;
 const MAX_ANTIGRAVITY_OUTPUT_TOKENS = 16384;
 
-// Full-jitter backoff config for 503 retries
-const BACKOFF_503_BASE_MS = 1000;
-const BACKOFF_503_CAP_MS = 15000;
+const ANTIGRAVITY_TRANSIENT_ERROR_PATTERNS = [
+  /high\s+traffic/i,
+  /agent\s+(execution\s+)?terminated\s+due\s+to\s+error/i,
+  /capacity/i,
+  /temporarily\s+unavailable/i,
+  /timeout/i,
+  /stream\s+(ended|closed|terminated|interrupted)/i,
+  /empty\s+response/i,
+];
+
+const ANTIGRAVITY_TRANSIENT_STATUSES = new Set([
+  HTTP_STATUS.SERVER_ERROR,
+  HTTP_STATUS.BAD_GATEWAY,
+  HTTP_STATUS.SERVICE_UNAVAILABLE,
+  HTTP_STATUS.GATEWAY_TIMEOUT,
+]);
+
+// Fields Google generateContent rejects (Claude/OpenAI/Qwen thinking fields set at body root by thinkingUnified.js)
+const ANTIGRAVITY_REQUEST_BLACKLIST = [
+  "output_config",
+  "thinking",
+  "reasoning_effort",
+  "reasoning",
+  "enable_thinking",
+  "thinking_budget",
+  "thinkingConfig",
+];
+
+// Strip blacklisted fields from an object (used for both body.request and top-level body)
+const stripBlacklisted = obj => {
+  for (const key of ANTIGRAVITY_REQUEST_BLACKLIST) delete obj[key];
+};
+
+// Image generation model name patterns
+const IMAGE_MODEL_PATTERNS = [
+  /image/i,
+  /imagen/i,
+  /image-generation/i,
+];
+
+// Detect if a model is an image generation model
+function isImageModel(model) {
+  if (!model) return false;
+  return IMAGE_MODEL_PATTERNS.some(p => p.test(model));
+}
+
+// Parse aspect ratio / resolution from model name suffixes
+// e.g. "gemini-3.1-flash-image-16x9" -> { aspectRatio: "16:9" }
+// e.g. "gemini-3.1-flash-image-1024x768" -> { aspectRatio: "4:3" }
+function parseImageConfig(model) {
+  const config = { aspectRatio: "1:1" };
+  const resMatch = model.match(/(\d+)x(\d+)$/);
+  if (resMatch) {
+    const w = parseInt(resMatch[1]);
+    const h = parseInt(resMatch[2]);
+    if (w <= 16 && h <= 16) {
+      config.aspectRatio = `${w}:${h}`;
+    } else {
+      // Resolution like 1024x768 — derive aspect ratio
+      const gcd = (a, b) => b ? gcd(b, a % b) : a;
+      const d = gcd(w, h);
+      config.aspectRatio = `${w/d}:${h/d}`;
+    }
+  }
+  return config;
+}
 
 export class AntigravityExecutor extends BaseExecutor {
   constructor() {
     super("antigravity", PROVIDERS.antigravity);
   }
 
+  getRetryConfig(credentials) {
+    return {
+      ...this.config.retry,
+      [HTTP_STATUS.SERVICE_UNAVAILABLE]: {
+        ...(typeof this.config.retry?.[HTTP_STATUS.SERVICE_UNAVAILABLE] === "object"
+          ? this.config.retry[HTTP_STATUS.SERVICE_UNAVAILABLE]
+          : {}),
+        attempts: credentials?.antigravity503RetryCount ?? DEFAULT_AG_503_RETRY_COUNT,
+      },
+    };
+  }
+
   buildUrl(model, stream, urlIndex = 0) {
     const baseUrls = this.getBaseUrls();
     const baseUrl = baseUrls[urlIndex] || baseUrls[0];
-    const action = stream ? "streamGenerateContent?alt=sse" : "generateContent";
+    // Image generation MUST use non-streaming generateContent
+    const forceNonStream = isImageModel(model);
+    const action = (stream && !forceNonStream) ? "streamGenerateContent?alt=sse" : "generateContent";
     return `${baseUrl}/v1internal:${action}`;
   }
 
+  // sessionId comes from transformRequest output; base.execute runs transformRequest before
+  // buildHeaders, so we read it from instance state cached there (fallback: explicit arg).
   buildHeaders(credentials, stream = true, sessionId = null) {
+    const sid = sessionId || this._lastSessionId;
     return {
       "Content-Type": "application/json",
       "Authorization": `Bearer ${credentials.accessToken}`,
       "User-Agent": this.config.headers?.["User-Agent"] || ANTIGRAVITY_HEADERS["User-Agent"],
       [INTERNAL_REQUEST_HEADER.name]: INTERNAL_REQUEST_HEADER.value,
-      ...(sessionId && { "X-Machine-Session-Id": sessionId }),
+      ...(sid && { "X-Machine-Session-Id": sid }),
       "Accept": stream ? "text/event-stream" : "application/json"
     };
   }
@@ -48,6 +130,53 @@ export class AntigravityExecutor extends BaseExecutor {
   transformRequest(model, body, stream, credentials) {
     const projectId = credentials?.projectId || this.generateProjectId();
 
+    // ─── Image generation: completely different request structure ───
+    if (isImageModel(model)) {
+      const imageConfig = parseImageConfig(model);
+      // Strip model name suffixes for the actual API model name
+      const cleanModel = model.replace(/-(\d+)x(\d+)$/, "");
+
+      // Build simplified contents — text-only, merge all user messages
+      const contents = [];
+      const srcContents = body.request?.contents || body.contents || [];
+      for (const c of srcContents) {
+        const textParts = (c.parts || []).filter(p => p.text !== undefined).map(p => ({ text: p.text }));
+        if (textParts.length > 0) {
+          contents.push({ role: c.role || "user", parts: textParts });
+        }
+      }
+
+      const sessionId = resolveSessionId({
+        headers: credentials?.rawHeaders,
+        body,
+        connectionId: credentials?.email || credentials?.connectionId,
+        scope: "antigravity",
+      });
+
+      this._lastSessionId = sessionId;
+
+      return {
+        project: projectId,
+        model: cleanModel,
+        userAgent: "antigravity",
+        requestType: "image_gen",
+        requestId: `agent-${crypto.randomUUID()}`,
+        request: {
+          contents,
+          generationConfig: {
+            temperature: 1.0,
+            topP: 0.95,
+            topK: 40,
+            maxOutputTokens: 8192,
+            imageConfig,
+          },
+          sessionId,
+          // No tools, no systemInstruction, no safetySettings for image gen
+        },
+      };
+    }
+
+    // ─── Standard (non-image) request ───
     // Fix contents for Claude models via Antigravity
     const contents = body.request?.contents?.map(c => {
       let role = c.role;
@@ -61,8 +190,19 @@ export class AntigravityExecutor extends BaseExecutor {
         if (p.thoughtSignature && !p.functionCall && !p.text) return false;
         return true;
       });
-      if (role !== c.role || parts?.length !== c.parts?.length) {
-        return { ...c, role, parts };
+      // Gemini 3+ rejects functionCall parts without thoughtSignature. Clients (Claude Code, IDE)
+      // don't persist thoughtSignature in their history, so backfill the default signature on any
+      // functionCall part that arrives without one.
+      const needsBackfill = parts?.some(p => p.functionCall && !p.thoughtSignature) ?? false;
+      if (role !== c.role || parts?.length !== c.parts?.length || needsBackfill) {
+        return {
+          ...c, role,
+          parts: needsBackfill
+            ? parts.map(p => (p.functionCall && !p.thoughtSignature)
+                ? { ...p, thoughtSignature: DEFAULT_THINKING_AG_SIGNATURE }
+                : p)
+            : parts,
+        };
       }
       return c;
     });
@@ -72,19 +212,28 @@ export class AntigravityExecutor extends BaseExecutor {
 
     if (tools && tools.length > 0) {
       // Merge all groups into a single functionDeclarations group (Gemini expects 1 group)
-      const allDeclarations = tools.flatMap(group =>
-        (group.functionDeclarations || []).map(fn => ({
-          ...fn,
-          name: sanitizeFunctionName(fn.name),
-          parameters: fn.parameters
-            ? cleanJSONSchemaForAntigravity(structuredClone(fn.parameters))
-            : { type: "object", properties: { reason: { type: "string", description: "Brief explanation" } }, required: ["reason"] }
-        }))
-      );
+      const seenToolNames = new Set();
+      const allDeclarations = [];
+      for (const group of tools) {
+        for (const fn of group.functionDeclarations || []) {
+          const name = sanitizeFunctionName(fn.name);
+          if (seenToolNames.has(name)) continue;
+          seenToolNames.add(name);
+          allDeclarations.push({
+            ...fn,
+            name,
+            parameters: fn.parameters
+              ? cleanJSONSchemaForAntigravity(structuredClone(fn.parameters))
+              : { type: "object", properties: { reason: { type: "string", description: "Brief explanation" } }, required: ["reason"] }
+          });
+        }
+      }
       tools = allDeclarations.length > 0 ? [{ functionDeclarations: allDeclarations }] : [];
     }
 
+    // Strip tools/toolConfig (handled separately) and blacklisted fields that Google rejects
     const { tools: _originalTools, toolConfig: _originalToolConfig, ...requestWithoutTools } = body.request || {};
+    stripBlacklisted(requestWithoutTools);
     const generationConfig = { ...(requestWithoutTools.generationConfig || {}) };
     if (generationConfig.maxOutputTokens > MAX_ANTIGRAVITY_OUTPUT_TOKENS) {
       generationConfig.maxOutputTokens = MAX_ANTIGRAVITY_OUTPUT_TOKENS;
@@ -95,11 +244,16 @@ export class AntigravityExecutor extends BaseExecutor {
       generationConfig,
       ...(contents && { contents }),
       ...(tools && { tools }),
-      sessionId: body.request?.sessionId || deriveSessionId(credentials?.email || credentials?.connectionId),
+      sessionId: body.request?.sessionId || resolveSessionId({ headers: credentials?.rawHeaders, body, connectionId: credentials?.email || credentials?.connectionId, scope: "antigravity" }),
       safetySettings: undefined,
       ...(tools?.length > 0 && { toolConfig: { functionCallingConfig: { mode: "VALIDATED" } } })
     };
 
+    // Strip blacklisted thinking fields from top-level body (set by thinkingUnified.js at root, not body.request)
+    stripBlacklisted(body);
+
+    this._lastSessionId = transformedRequest.sessionId; // cached for buildHeaders (base.execute order)
+
     return {
       ...body,
       project: projectId,
@@ -200,117 +354,49 @@ export class AntigravityExecutor extends BaseExecutor {
     return totalMs > 0 ? totalMs : null;
   }
 
-  async execute({ model, body, stream, credentials, signal, log, proxyOptions = null }) {
-    const fallbackCount = this.getFallbackCount();
-    let lastError = null;
-    let lastStatus = 0;
-    const MAX_AUTO_RETRIES = 3;
-    const MAX_RETRY_AFTER_RETRIES = 3;
-    // 503 retry count: resolved by auth layer (per-account → global → default)
-    const MAX_503_RETRIES = credentials?.antigravity503RetryCount ?? DEFAULT_AG_503_RETRY_COUNT;
-    const retryAttemptsByUrl = {}; // Track retry attempts per URL
-    const retryAfterAttemptsByUrl = {}; // Track Retry-After retries per URL
-    const retry503AttemptsByUrl = {}; // Track 503-specific retries per URL
-
-    for (let urlIndex = 0; urlIndex < fallbackCount; urlIndex++) {
-      const url = this.buildUrl(model, stream, urlIndex);
-      const transformedBody = this.transformRequest(model, body, stream, credentials);
-      const sessionId = transformedBody.request?.sessionId;
-      const headers = this.buildHeaders(credentials, stream, sessionId);
-
-      // Initialize retry counters for this URL
-      if (!retryAttemptsByUrl[urlIndex]) {
-        retryAttemptsByUrl[urlIndex] = 0;
-      }
-      if (!retryAfterAttemptsByUrl[urlIndex]) {
-        retryAfterAttemptsByUrl[urlIndex] = 0;
-      }
-      if (!retry503AttemptsByUrl[urlIndex]) {
-        retry503AttemptsByUrl[urlIndex] = 0;
-      }
+  extractErrorMessage(errorJson, bodyText = "") {
+    return [
+      errorJson?.error?.message,
+      errorJson?.message,
+      errorJson?.error,
+      bodyText,
+    ].filter(Boolean).map(v => typeof v === "string" ? v : JSON.stringify(v)).join("\n");
+  }
 
-      try {
-        const response = await proxyAwareFetch(url, {
-          method: "POST",
-          headers,
-          body: JSON.stringify(transformedBody),
-          signal
-        }, proxyOptions);
-
-        if (response.status === HTTP_STATUS.RATE_LIMITED || response.status === HTTP_STATUS.SERVICE_UNAVAILABLE) {
-          // 503-specific in-place retry with exponential backoff (before Retry-After check)
-          if (response.status === HTTP_STATUS.SERVICE_UNAVAILABLE && retry503AttemptsByUrl[urlIndex] < MAX_503_RETRIES) {
-            retry503AttemptsByUrl[urlIndex]++;
-            // Full jitter: random in [0, min(cap, base * 2^attempt)]
-            // Prevents thundering herd when multiple requests hit 503 simultaneously
-            const expCeiling = Math.min(BACKOFF_503_CAP_MS, BACKOFF_503_BASE_MS * Math.pow(2, retry503AttemptsByUrl[urlIndex] - 1));
-            const backoffMs = Math.floor(Math.random() * expCeiling);
-            log?.debug?.("RETRY", `503 retry ${retry503AttemptsByUrl[urlIndex]}/${MAX_503_RETRIES} after ${(backoffMs / 1000).toFixed(2)}s (jitter, ceil=${expCeiling / 1000}s)`);
-            await new Promise(resolve => setTimeout(resolve, backoffMs));
-            urlIndex--;
-            continue;
-          }
-
-          // Try to get retry time from headers first
-          let retryMs = this.parseRetryHeaders(response.headers);
-
-          // If no retry time in headers, try to parse from error message body
-          if (!retryMs) {
-            try {
-              const errorBody = await response.clone().text();
-              const errorJson = JSON.parse(errorBody);
-              const errorMessage = errorJson?.error?.message || errorJson?.message || "";
-              retryMs = this.parseRetryFromErrorMessage(errorMessage);
-            } catch (e) {
-              // Ignore parse errors, will fall back to exponential backoff
-            }
-          }
-
-          if (retryMs && retryMs <= MAX_RETRY_AFTER_MS && retryAfterAttemptsByUrl[urlIndex] < MAX_RETRY_AFTER_RETRIES) {
-            retryAfterAttemptsByUrl[urlIndex]++;
-            log?.debug?.("RETRY", `${response.status} with Retry-After: ${Math.ceil(retryMs / 1000)}s, waiting... (${retryAfterAttemptsByUrl[urlIndex]}/${MAX_RETRY_AFTER_RETRIES})`);
-            await new Promise(resolve => setTimeout(resolve, retryMs));
-            urlIndex--;
-            continue;
-          }
-
-          // Auto retry only for 429 when retryMs is 0 or undefined
-          if (response.status === HTTP_STATUS.RATE_LIMITED && (!retryMs || retryMs === 0) && retryAttemptsByUrl[urlIndex] < MAX_AUTO_RETRIES) {
-            retryAttemptsByUrl[urlIndex]++;
-            // Exponential backoff: 2s, 4s, 8s...
-            const backoffMs = Math.min(1000 * (2 ** retryAttemptsByUrl[urlIndex]), MAX_RETRY_AFTER_MS);
-            log?.debug?.("RETRY", `429 auto retry ${retryAttemptsByUrl[urlIndex]}/${MAX_AUTO_RETRIES} after ${backoffMs / 1000}s`);
-            await new Promise(resolve => setTimeout(resolve, backoffMs));
-            urlIndex--;
-            continue;
-          }
-
-          log?.debug?.("RETRY", `${response.status}, Retry-After ${retryMs ? `too long (${Math.ceil(retryMs / 1000)}s)` : 'missing'}, trying fallback`);
-          lastStatus = response.status;
-
-          if (urlIndex + 1 < fallbackCount) {
-            continue;
-          }
-        }
+  isTransientAntigravityError(status, message) {
+    if (status === HTTP_STATUS.RATE_LIMITED) return true;
+    if (ANTIGRAVITY_TRANSIENT_STATUSES.has(status)) return true;
+    return ANTIGRAVITY_TRANSIENT_ERROR_PATTERNS.some(pattern => pattern.test(message || ""));
+  }
 
-        if (this.shouldRetry(response.status, urlIndex)) {
-          log?.debug?.("RETRY", `${response.status} on ${url}, trying fallback ${urlIndex + 1}`);
-          lastStatus = response.status;
-          continue;
-        }
+  // Hook called by BaseExecutor.tryRetry: derive delay from Retry-After (header → body),
+  // cap at MAX_RETRY_AFTER_MS, else retry transient Antigravity failures with backoff.
+  // Return false to veto (fallback URL / final error).
+  async computeRetryDelay(response, attempt) {
+    let bodyText = "";
+    let errorJson = null;
+    let retryMs = this.parseRetryHeaders(response.headers);
 
-        return { response, url, headers, transformedBody };
-      } catch (error) {
-        lastError = error;
-        if (urlIndex + 1 < fallbackCount) {
-          log?.debug?.("RETRY", `Error on ${url}, trying fallback ${urlIndex + 1}`);
-          continue;
-        }
-        throw error;
-      }
+    try {
+      bodyText = await response.clone().text();
+      errorJson = bodyText ? JSON.parse(bodyText) : null;
+    } catch {
+      // ignore parse errors → fall through to status/message based retry
+    }
+
+    const errorMessage = this.extractErrorMessage(errorJson, bodyText);
+
+    if (!retryMs) {
+      retryMs = this.parseRetryFromErrorMessage(errorMessage);
     }
+    if (retryMs) return retryMs <= MAX_RETRY_AFTER_MS ? retryMs : false;
+
+    if (!this.isTransientAntigravityError(response.status, errorMessage)) return false;
 
-    throw lastError || new Error(`All ${fallbackCount} URLs failed with status ${lastStatus}`);
+    const cap = response.status === HTTP_STATUS.RATE_LIMITED
+      ? MAX_RETRY_AFTER_MS
+      : ANTIGRAVITY_TRANSIENT_RETRY_MAX_MS;
+    return Math.min(1000 * (2 ** attempt), cap); // exponential backoff
   }
 
   /**

package/app/open-sse/executors/base.js

@@ -1,6 +1,8 @@
 import { HTTP_STATUS, RETRY_CONFIG, DEFAULT_RETRY_CONFIG, resolveRetryEntry, FETCH_CONNECT_TIMEOUT_MS } from "../config/runtimeConfig.js";
+import { shouldRefreshCredentials } from "../services/oauthCredentialManager.js";
 import { proxyAwareFetch } from "../utils/proxyFetch.js";
 import { dbg } from "../utils/debugLog.js";
+import { ANTHROPIC_API_VERSION, OPENAI_COMPAT_BASE, ANTHROPIC_COMPAT_BASE } from "../providers/shared.js";
 
 /**
  * BaseExecutor - Base class for provider executors
@@ -26,13 +28,13 @@ export class BaseExecutor {
 
   buildUrl(model, stream, urlIndex = 0, credentials = null) {
     if (this.provider?.startsWith?.("openai-compatible-")) {
-      const baseUrl = credentials?.providerSpecificData?.baseUrl || "https://api.openai.com/v1";
+      const baseUrl = credentials?.providerSpecificData?.baseUrl || OPENAI_COMPAT_BASE;
       const normalized = baseUrl.replace(/\/$/, "");
       const path = this.provider.includes("responses") ? "/responses" : "/chat/completions";
       return `${normalized}${path}`;
     }
     if (this.provider?.startsWith?.("anthropic-compatible-")) {
-      const baseUrl = credentials?.providerSpecificData?.baseUrl || "https://api.anthropic.com/v1";
+      const baseUrl = credentials?.providerSpecificData?.baseUrl || ANTHROPIC_COMPAT_BASE;
       const normalized = baseUrl.replace(/\/$/, "");
       return `${normalized}/messages`;
     }
@@ -54,7 +56,7 @@ export class BaseExecutor {
         headers["Authorization"] = `Bearer ${credentials.accessToken}`;
       }
       if (!headers["anthropic-version"]) {
-        headers["anthropic-version"] = "2023-06-01";
+        headers["anthropic-version"] = ANTHROPIC_API_VERSION;
       }
     } else {
       // Standard Bearer token auth for other providers
@@ -87,9 +89,7 @@ export class BaseExecutor {
   }
 
   needsRefresh(credentials) {
-    if (!credentials.expiresAt) return false;
-    const expiresAtMs = new Date(credentials.expiresAt).getTime();
-    return expiresAtMs - Date.now() < 5 * 60 * 1000;
+    return shouldRefreshCredentials(this.provider, credentials);
   }
 
   parseError(response, bodyText) {
@@ -103,15 +103,24 @@ export class BaseExecutor {
     const retryAttemptsByUrl = {};
 
     // Merge default retry config with provider-specific config
-    const retryConfig = { ...DEFAULT_RETRY_CONFIG, ...this.config.retry };
+    const providerRetryConfig = this.getRetryConfig?.(credentials) || this.config.retry;
+    const retryConfig = { ...DEFAULT_RETRY_CONFIG, ...providerRetryConfig };
 
     // Schedule retry via retryConfig[statusKey]. Returns true when caller should `urlIndex--; continue`
-    const tryRetry = async (urlIndex, statusKey, reason) => {
+    // response (optional) lets a subclass hook compute a dynamic delay (e.g. antigravity Retry-After).
+    const tryRetry = async (urlIndex, statusKey, reason, response = null) => {
       const { attempts, delayMs } = resolveRetryEntry(retryConfig[statusKey]);
       if (attempts <= 0 || retryAttemptsByUrl[urlIndex] >= attempts) return false;
+      // Hook: subclass may derive delay from the response (headers/body). null → skip retry, use fallback.
+      let waitMs = delayMs;
+      if (response && this.computeRetryDelay) {
+        const dynamic = await this.computeRetryDelay(response, retryAttemptsByUrl[urlIndex] + 1, delayMs);
+        if (dynamic === false) return false; // hook vetoes retry (e.g. Retry-After too long)
+        if (dynamic != null) waitMs = dynamic;
+      }
       retryAttemptsByUrl[urlIndex]++;
-      log?.debug?.("RETRY", `${reason} retry ${retryAttemptsByUrl[urlIndex]}/${attempts} after ${delayMs / 1000}s`);
-      await new Promise(resolve => setTimeout(resolve, delayMs));
+      log?.debug?.("RETRY", `${reason} retry ${retryAttemptsByUrl[urlIndex]}/${attempts} after ${waitMs / 1000}s`);
+      await new Promise(resolve => setTimeout(resolve, waitMs));
       return true;
     };
 
@@ -122,15 +131,16 @@ export class BaseExecutor {
 
       if (!retryAttemptsByUrl[urlIndex]) retryAttemptsByUrl[urlIndex] = 0;
 
-      // Abort if upstream doesn't return response headers within FETCH_CONNECT_TIMEOUT_MS
+      // Abort if upstream doesn't return response headers within connection timeout
       const connectCtrl = new AbortController();
-      const connectTimer = setTimeout(() => connectCtrl.abort(new Error("fetch connect timeout")), FETCH_CONNECT_TIMEOUT_MS);
+      const timeoutMs = this.config?.timeoutMs || FETCH_CONNECT_TIMEOUT_MS;
+      const connectTimer = setTimeout(() => connectCtrl.abort(new Error("fetch connect timeout")), timeoutMs);
       const mergedSignal = signal ? AbortSignal.any([signal, connectCtrl.signal]) : connectCtrl.signal;
 
       try {
         const bodyStr = JSON.stringify(transformedBody);
         const fetchT0 = Date.now();
-        dbg("FETCH", `${this.provider.toUpperCase()} → ${url} | body=${bodyStr.length}B | connectTimeout=${FETCH_CONNECT_TIMEOUT_MS}ms`);
+        dbg("FETCH", `${this.provider.toUpperCase()} → ${url} | body=${bodyStr.length}B | connectTimeout=${timeoutMs}ms`);
         const response = await proxyAwareFetch(url, {
           method: "POST",
           headers,
@@ -142,7 +152,7 @@ export class BaseExecutor {
         const cl = response.headers?.get?.("content-length") || "?";
         dbg("FETCH", `${this.provider.toUpperCase()} ← ${response.status} | ttft=${Date.now() - fetchT0}ms | ct=${ct} | cl=${cl}`);
 
-        if (await tryRetry(urlIndex, response.status, `status ${response.status}`)) { urlIndex--; continue; }
+        if (await tryRetry(urlIndex, response.status, `status ${response.status}`, response)) { urlIndex--; continue; }
 
         if (this.shouldRetry(response.status, urlIndex)) {
           log?.debug?.("RETRY", `${response.status} on ${url}, trying fallback ${urlIndex + 1}`);

package/app/open-sse/executors/codebuddy-cn.js

@@ -0,0 +1,40 @@
+import { DefaultExecutor } from "./default.js";
+
+/**
+ * CodeBuddyExecutor — talks to https://copilot.tencent.com/v2/chat/completions
+ *
+ * CodeBuddy is OpenAI-compatible but rejects non-stream chat requests
+ * (HTTP 400, code 11101 "Non-stream chat request is currently not supported").
+ * The same-format (openai→openai) translator path leaves body.stream as the
+ * client sent it, so we force it true here — 9router still re-aggregates the
+ * SSE into a JSON response for non-streaming clients.
+ */
+export class CodeBuddyExecutor extends DefaultExecutor {
+  constructor() {
+    super("codebuddy-cn");
+  }
+
+  transformRequest(model, body, stream, credentials) {
+    const transformed = super.transformRequest(model, body, stream, credentials);
+    transformed.stream = true;
+
+    // CodeBuddy only surfaces model reasoning when the request carries the CLI's
+    // OpenAI-style params: reasoning_effort + reasoning_summary:"auto". 9router's
+    // thinking pipeline sets reasoning_effort only when the client asks, and never
+    // sets reasoning_summary — so reasoning never shows. Mirror the CLI here.
+    const eff = transformed.reasoning_effort;
+    if (eff === "none" || eff === "off") {
+      delete transformed.reasoning_effort; // gateway has no "none" — just omit
+    } else if (eff) {
+      // Client explicitly asked for reasoning — mirror the CLI's reasoning_summary
+      // so CodeBuddy surfaces the model's reasoning.
+      transformed.reasoning_summary = "auto";
+    }
+    // No reasoning requested: leave both unset. Forcing reasoning_effort:"medium"
+    // + reasoning_summary on plain requests makes CodeBuddy trip its content
+    // filter and return an error (#2071).
+    return transformed;
+  }
+}
+
+export default CodeBuddyExecutor;