swayrouter 1.0.8 → 1.3.0

package/src/routes/api/settings/route.js

@@ -1,6 +1,6 @@
 import { NextResponse } from "next/server";
-import { getSettings, updateSettings } from "@/lib/localDb";
-import { stripRemovedSettings } from "@/lib/db/repos/settingsRepo.js";
+import { getSettings, updateSettings } from "@/lib/localDb";
+import { stripRemovedSettings, validateFallbackStrategies } from "@/lib/db/repos/settingsRepo.js";
 import { resetComboRotation } from "open-sse/services/combo.js";
 import bcrypt from "bcryptjs";
 import { cookies } from "next/headers";
@@ -21,13 +21,13 @@ const PROTECTED_SETTING_KEYS = ["password", "dbPath", "tunnelEnabled", "tunnelUr
 
 const CLI_MODEL_MAPPING_KEYS = new Set(["enabled", "entries"]);
 
-function sanitizeSettingsForResponse(settings) {
-  const {
-    password: _password,
-    dbPath: _dbPath,
-    ...safeSettings
-  } = stripRemovedSettings(settings);
-  return safeSettings;
+function sanitizeSettingsForResponse(settings) {
+  const {
+    password: _password,
+    dbPath: _dbPath,
+    ...safeSettings
+  } = stripRemovedSettings(settings);
+  return safeSettings;
 }
 
 export const __test__ = { sanitizeSettingsForResponse };
@@ -86,6 +86,15 @@ export async function PATCH(request) {
 
     for (const key of PROTECTED_SETTING_KEYS) delete body[key];
 
+    try {
+      validateFallbackStrategies(body);
+    } catch (error) {
+      if (error?.code === "INVALID_FALLBACK_STRATEGY") {
+        return NextResponse.json({ error: error.message }, { status: 400 });
+      }
+      throw error;
+    }
+
     const currentSettings = await getSettings();
     if (
       currentSettings.tunnelEnabled === true &&
@@ -126,31 +135,31 @@ export async function PATCH(request) {
       body.cliModelMappings = sanitizeCliModelMappings(body.cliModelMappings);
     }
 
-    if (Object.prototype.hasOwnProperty.call(body, "profileName")) {
-      body.profileName = typeof body.profileName === "string"
-        ? body.profileName.trim().replace(/\s+/g, " ").slice(0, 40)
-        : "";
-    }
-
-    if (Object.prototype.hasOwnProperty.call(body, "currency")) {
-      body.currency = body.currency === "IDR" ? "IDR" : "USD";
-    }
-
-    if (Object.prototype.hasOwnProperty.call(body, "profileAvatar")) {
-      const avatar = typeof body.profileAvatar === "string" ? body.profileAvatar.trim() : "";
-      const isValidAvatar = avatar === ""
-        || (avatar.length <= 700_000
-          && /^data:image\/(?:png|jpeg|webp|gif);base64,[A-Za-z0-9+/=\r\n]+$/.test(avatar));
-      if (!isValidAvatar) {
-        return NextResponse.json(
-          { error: "Profile photo must be a PNG, JPG, WEBP, or GIF up to 512 KB" },
-          { status: 400 },
-        );
-      }
-      body.profileAvatar = avatar;
-    }
-
-    const settings = await updateSettings(body);
+    if (Object.prototype.hasOwnProperty.call(body, "profileName")) {
+      body.profileName = typeof body.profileName === "string"
+        ? body.profileName.trim().replace(/\s+/g, " ").slice(0, 40)
+        : "";
+    }
+
+    if (Object.prototype.hasOwnProperty.call(body, "currency")) {
+      body.currency = body.currency === "IDR" ? "IDR" : "USD";
+    }
+
+    if (Object.prototype.hasOwnProperty.call(body, "profileAvatar")) {
+      const avatar = typeof body.profileAvatar === "string" ? body.profileAvatar.trim() : "";
+      const isValidAvatar = avatar === ""
+        || (avatar.length <= 700_000
+          && /^data:image\/(?:png|jpeg|webp|gif);base64,[A-Za-z0-9+/=\r\n]+$/.test(avatar));
+      if (!isValidAvatar) {
+        return NextResponse.json(
+          { error: "Profile photo must be a PNG, JPG, WEBP, or GIF up to 512 KB" },
+          { status: 400 },
+        );
+      }
+      body.profileAvatar = avatar;
+    }
+
+    const settings = await updateSettings(body);
 
     if (forcedPasswordChange) {
 

package/src/services/accountFallback.js

@@ -3,13 +3,15 @@ import {
   BACKOFF_CONFIG,
   TRANSIENT_COOLDOWN_MS,
   classifyError,
-  ERROR_TIERS,
-  isDeprioitized,
-  isCancellationStatus,
-  isNonRetryableRequestStatus,
-} from "../config/errorConfig.js";
+  ERROR_TIERS,
+  isDeprioitized,
+  isCancellationStatus,
+  isNonRetryableRequestStatus,
+  isQuotaExhaustionError,
+} from "../config/errorConfig.js";
 
 import { fnv1a } from "@/lib/security/payloadCapture.js";
+import { extractClientSessionId } from "../utils/sessionManager.js";
 
 export function getQuotaCooldown(backoffLevel = 0) {
   const level = Math.max(0, backoffLevel - 1);
@@ -18,7 +20,12 @@ export function getQuotaCooldown(backoffLevel = 0) {
 }
 
 export function checkFallbackError(status, errorText, backoffLevel = 0) {
-  if (isNonRetryableRequestStatus(status) || isCancellationStatus(status)) {
+  if (isQuotaExhaustionError(errorText)) {
+    const newLevel = Math.min(backoffLevel + 1, BACKOFF_CONFIG.maxLevel);
+    return { shouldFallback: true, cooldownMs: getQuotaCooldown(newLevel), newBackoffLevel: newLevel };
+  }
+
+  if (isNonRetryableRequestStatus(status) || isCancellationStatus(status)) {
     return { shouldFallback: false, cooldownMs: 0, action: "final", newBackoffLevel: backoffLevel };
   }
 
@@ -193,25 +200,52 @@ export function pickCacheAffine(connections, { cacheKey, excludeIds } = {}) {
   const all = connections || [];
   if (all.length === 0) return { connection: null, affinity: "miss" };
   const excluded = excludeIds instanceof Set ? excludeIds : (excludeIds ? new Set(excludeIds) : null);
+  const isExcluded = (c) => !!(excluded && c && excluded.has(c.id));
+  const prio = (c) => (c && c.priority != null ? c.priority : 999);
+
+  // Stable copy ordered by ascending priority (lower number = higher priority).
+  const ordered = all.slice().sort((a, b) => prio(a) - prio(b));
 
   if (!cacheKey) {
 
-    const firstAvail = excluded ? all.find((c) => !excluded.has(c.id)) : all[0];
+    const firstAvail = ordered.find((c) => !isExcluded(c));
     return { connection: firstAvail || null, affinity: "miss" };
   }
 
-  const n = all.length;
-  const bucket = fnv1a(cacheKey) % n;
-  const primary = all[bucket] || null;
-
-  if (primary && (!excluded || !excluded.has(primary.id))) {
-    return { connection: primary, affinity: "hit" };
+  // Group into priority tiers, preserving stable order within each tier.
+  const tiers = [];
+  for (const c of ordered) {
+    const p = prio(c);
+    const last = tiers[tiers.length - 1];
+    if (!last || last.priority !== p) tiers.push({ priority: p, members: [c] });
+    else last.members.push(c);
   }
 
-  for (let step = 1; step <= n; step++) {
-    const cand = all[(bucket + step) % n];
-    if (!cand || (excluded && excluded.has(cand.id))) continue;
-    return { connection: cand, affinity: "rotate" };
+  // Highest-priority tier with capacity wins; affinity is "hit" only when the
+  // top tier's hashed primary is itself available, otherwise "rotate".
+  let rotated = false;
+  for (const tier of tiers) {
+    const members = tier.members;
+    if (members.length === 0) continue;
+    if (members.every((c) => isExcluded(c))) {
+      // Whole tier excluded → fall through to the next tier.
+      rotated = true;
+      continue;
+    }
+
+    const bucket = fnv1a(cacheKey) % members.length;
+    const primary = members[bucket];
+    if (!isExcluded(primary)) {
+      return { connection: primary, affinity: rotated ? "rotate" : "hit" };
+    }
+
+    for (let step = 1; step < members.length; step++) {
+      const cand = members[(bucket + step) % members.length];
+      if (!isExcluded(cand)) return { connection: cand, affinity: "rotate" };
+    }
+
+    // Defensive: tier had a non-excluded member but the walk missed it.
+    rotated = true;
   }
 
   return { connection: null, affinity: "rotate" };
@@ -252,3 +286,30 @@ export function extractCacheKey(body) {
   if (typeof body.input === "string" && body.input) return body.input.slice(0, 256);
   return null;
 }
+
+// Resolve a stable cache-affine key for a request.
+// Priority:
+//   (a) an EXPLICIT client-provided session/conversation identity (headers, or
+//       body prompt_cache_key / session_id / conversation_id / metadata.user_id).
+//       An explicit id means "one conversation" → pin it to one account so every
+//       multi-turn request reuses that provider account's prompt cache.
+//       We deliberately ignore the random/derived session fallback (generateBinaryStyleId
+//       / deriveSessionId) here — it is unstable and would defeat affinity.
+//   (b) otherwise fall back to the prompt-prefix text (extractCacheKey), which is
+//       the best available proxy and also shares a common system prompt across clients.
+// Bound to <=256 chars. Format-agnostic (OpenAI messages, Gemini contents, plain prompt/input).
+/**
+ * @param {{ headers?: Record<string, string>|null, body?: unknown, connectionId?: string|null, scope?: string }} [opts]
+ * @returns {string|null}
+ */
+export function deriveCacheKey({ headers, body, connectionId, scope = "" } = {}) {
+  try {
+    // includeRequestId:false — x-client-request-id is PER-REQUEST, not per-conversation;
+    // letting it through would make the key change every call and defeat affinity.
+    const explicit = extractClientSessionId(headers, body, scope, { includeRequestId: false });
+    if (explicit) return `session:${explicit}`.slice(0, 256);
+  } catch {
+
+  }
+  return extractCacheKey(body);
+}

package/src/sse/handlers/chat.js

@@ -7,7 +7,7 @@ import {
   extractApiKey,
   isValidApiKey,
 } from "../services/auth.js";
-import { extractCacheKey } from "../../services/accountFallback.js";
+import { deriveCacheKey } from "../../services/accountFallback.js";
 import { getSettings, getProviderConnectionById } from "@/lib/localDb";
 import { getModelInfo, getComboModels } from "../services/model.js";
 import { handleChatCore } from "../../handlers/chatCore.js";
@@ -22,13 +22,13 @@ import { updateProviderCredentials, checkAndRefreshToken } from "../services/tok
 import { getProjectIdForConnection } from "../../services/projectId.js";
 import { getConsistentMachineId } from "@/shared/utils/machineId";
 import { isCallerCancellation } from "../services/requestCancellation.js";
-import { env } from "@/lib/env";
-
-const INTERNAL_CLI_TOKEN_HEADER = "x-swayrouter-cli-token";
-const INTERNAL_CONNECTION_HEADER = "x-swayrouter-connection-id";
-const INTERNAL_CLI_TOKEN_SALT = "swayrouter-cli-auth";
-
-function attachAccountLease(response, release) {
+import { env } from "@/lib/env";
+
+const INTERNAL_CLI_TOKEN_HEADER = "x-swayrouter-cli-token";
+const INTERNAL_CONNECTION_HEADER = "x-swayrouter-connection-id";
+const INTERNAL_CLI_TOKEN_SALT = "swayrouter-cli-auth";
+
+function attachAccountLease(response, release) {
   if (!response || typeof release !== "function") return response;
 
   let released = false;
@@ -240,11 +240,11 @@ export async function handleChat(request, clientRawRequest = null) {
   return handleSingleModelChat(body, modelStr, clientRawRequest, request, apiKey, preferredConnectionId, request?.signal);
 }
 
-async function handleSingleModelChat(body, modelStr, clientRawRequest = null, request = null, apiKey = null, preferredConnectionId = null, requestSignal = null) {
-  requestSignal ||= request?.signal || null;
-  if (requestSignal?.aborted) return errorResponse(499, "Request aborted");
-  const deadlineAt = Date.now() + env.upstreamDeadlineMs;
-  const incomingClient = clientRawRequest?.headers
+async function handleSingleModelChat(body, modelStr, clientRawRequest = null, request = null, apiKey = null, preferredConnectionId = null, requestSignal = null) {
+  requestSignal ||= request?.signal || null;
+  if (requestSignal?.aborted) return errorResponse(499, "Request aborted");
+  const deadlineAt = Date.now() + env.upstreamDeadlineMs;
+  const incomingClient = clientRawRequest?.headers
     ? Object.fromEntries(clientRawRequest.headers.entries?.() || [])
     : {};
   const settings = await getSettings();
@@ -312,17 +312,23 @@ async function handleSingleModelChat(body, modelStr, clientRawRequest = null, re
 
   const excludeConnectionIds = new Set();
   let lastError = null;
-  let lastStatus = null;
-
-  const cacheKey = extractCacheKey(body);
-  let retryCount = 0;
-  let accountAttempts = 0;
-
-  while (accountAttempts < env.maxAccountAttempts) {
-    if (Date.now() >= deadlineAt) {
-      return errorResponse(HTTP_STATUS.GATEWAY_TIMEOUT, "Sway Router upstream deadline exceeded");
-    }
-
+  let lastStatus = null;
+
+  const requestHeaders = clientRawRequest?.headers || Object.fromEntries(request.headers.entries());
+  const cacheKey = deriveCacheKey({
+    headers: requestHeaders,
+    body,
+    connectionId: preferredConnectionId || null,
+    scope: provider || "",
+  });
+  let retryCount = 0;
+  let accountAttempts = 0;
+
+  while (accountAttempts < env.maxAccountAttempts) {
+    if (Date.now() >= deadlineAt) {
+      return errorResponse(HTTP_STATUS.GATEWAY_TIMEOUT, "Sway Router upstream deadline exceeded");
+    }
+
     const credentials = await getProviderCredentials(provider, excludeConnectionIds, model, {
       cacheKey,
       preferredConnectionId,
@@ -354,10 +360,10 @@ async function handleSingleModelChat(body, modelStr, clientRawRequest = null, re
         credentials.retryAfter,
         credentials.retryAfterHuman || "retry shortly",
       );
-    }
-
-    accountAttempts += 1;
-    const releaseAccountSlot = credentials.releaseAccountSlot;
+    }
+
+    accountAttempts += 1;
+    const releaseAccountSlot = credentials.releaseAccountSlot;
     let result;
     try {
       const refreshedCredentials = await checkAndRefreshToken(provider, credentials);
@@ -375,7 +381,7 @@ async function handleSingleModelChat(body, modelStr, clientRawRequest = null, re
       const providerThinking = (chatSettings.providerThinking || {})[provider] || null;
       result = await handleChatCore({
         body: { ...body, model: `${provider}/${model}` },
-        modelInfo: { provider, model, modelCapabilities: modelInfo.modelCapabilities },
+        modelInfo: { provider, model, modelCapabilities: modelInfo.modelCapabilities },
         credentials: refreshedCredentials,
         log,
         clientRawRequest,
@@ -388,9 +394,9 @@ async function handleSingleModelChat(body, modelStr, clientRawRequest = null, re
         cavemanLevel: chatSettings.cavemanLevel || "full",
         ponytailEnabled: !!chatSettings.ponytailEnabled,
         ponytailLevel: chatSettings.ponytailLevel || "full",
-        providerThinking,
-        requestSignal,
-        deadlineAt,
+        providerThinking,
+        requestSignal,
+        deadlineAt,
 
         sourceFormatOverride: request?.url ? detectFormatByEndpoint(new URL(request.url).pathname, body) : null,
         onCredentialsRefreshed: async (newCreds) => {
@@ -414,33 +420,33 @@ async function handleSingleModelChat(body, modelStr, clientRawRequest = null, re
 
     releaseAccountSlot?.();
 
-    if (isCallerCancellation({ requestSignal, status: result.status })) {
-      return result.response;
-    }
-
-    if (Date.now() >= deadlineAt) {
-      return result.response.status === HTTP_STATUS.GATEWAY_TIMEOUT
-        ? result.response
-        : errorResponse(HTTP_STATUS.GATEWAY_TIMEOUT, "Sway Router upstream deadline exceeded");
-    }
+    if (isCallerCancellation({ requestSignal, status: result.status })) {
+      return result.response;
+    }
+
+    if (Date.now() >= deadlineAt) {
+      return result.response.status === HTTP_STATUS.GATEWAY_TIMEOUT
+        ? result.response
+        : errorResponse(HTTP_STATUS.GATEWAY_TIMEOUT, "Sway Router upstream deadline exceeded");
+    }
 
     const { shouldFallback } = await markAccountUnavailable(credentials.connectionId, result.status, result.error, provider, model, result.resetsAtMs, { retryCount });
 
-    if (shouldFallback) {
-      log.warn("FALLBACK", `⇄ ACC:${credentials.connectionName} UNAVAILABLE (${result.status}) → NEXT ACCOUNT`);
-      excludeConnectionIds.add(credentials.connectionId);
-      lastError = result.error;
-      lastStatus = result.status;
-      retryCount += 1;
-      if (accountAttempts >= env.maxAccountAttempts) {
-        log.warn("FALLBACK", `[${provider}/${model}] account attempt limit reached (${env.maxAccountAttempts})`);
-        return errorResponse(HTTP_STATUS.SERVICE_UNAVAILABLE, "Provider account fallback limit reached; retry shortly");
-      }
-      continue;
+    if (shouldFallback) {
+      log.warn("FALLBACK", `⇄ ACC:${credentials.connectionName} UNAVAILABLE (${result.status}) → NEXT ACCOUNT`);
+      excludeConnectionIds.add(credentials.connectionId);
+      lastError = result.error;
+      lastStatus = result.status;
+      retryCount += 1;
+      if (accountAttempts >= env.maxAccountAttempts) {
+        log.warn("FALLBACK", `[${provider}/${model}] account attempt limit reached (${env.maxAccountAttempts})`);
+        return errorResponse(HTTP_STATUS.SERVICE_UNAVAILABLE, "Provider account fallback limit reached; retry shortly");
+      }
+      continue;
     }
 
-    return result.response;
-  }
-
-  return errorResponse(HTTP_STATUS.SERVICE_UNAVAILABLE, "Provider account fallback limit reached; retry shortly");
-}
+    return result.response;
+  }
+
+  return errorResponse(HTTP_STATUS.SERVICE_UNAVAILABLE, "Provider account fallback limit reached; retry shortly");
+}

package/src/sse/services/auth.js

@@ -7,7 +7,7 @@ import {
   isDeprioritized, sortWithDeprioritization, pickCacheAffine,
   DEPRIORITIZE_FIELD,
 } from "../../services/accountFallback.js";
-import { MAX_RATE_LIMIT_COOLDOWN_MS, classifyError, ERROR_TIERS } from "../../config/errorConfig.js";
+import { MAX_RATE_LIMIT_COOLDOWN_MS, classifyError, ERROR_TIERS, isQuotaExhaustionError } from "../../config/errorConfig.js";
 import { HTTP_STATUS } from "../../config/runtimeConfig.js";
 import { recordLock, clearLock, clearAllLocksForConnection } from "@/lib/db/repos/accountModelLocksRepo.js";
 import { accountFallbackDecisionsTotal, cacheAffineSelectionsTotal } from "@/observability/metrics.js";
@@ -40,6 +40,13 @@ function githubMonthlyResetMs(status, errorText, provider) {
   return Date.UTC(now.getUTCFullYear(), now.getUTCMonth() + 1, 1);
 }
 
+/**
+ * @param {string} provider
+ * @param {Set<string>|string|null} [excludeConnectionIds]
+ * @param {string|null} [model]
+ * @param {{ cacheKey?: string|null, promptPrefix?: string|null, preferredConnectionId?: string|null, signal?: AbortSignal|null, [key: string]: unknown }} [options]
+ * @returns {Promise<Record<string, unknown>|null>}
+ */
 export async function getProviderCredentials(provider, excludeConnectionIds = null, model = null, options = {}) {
 
   const excludeSet = excludeConnectionIds instanceof Set
@@ -398,13 +405,18 @@ export async function markAccountUnavailable(connectionId, status, errorText, pr
   if (!shouldFallback) return { shouldFallback: false, cooldownMs: 0 };
 
   const reason = typeof errorText === "string" ? errorText.slice(0, 100) : "Provider error";
-  const lockUpdate = buildModelLockUpdate(githubResetAtMs ? null : model, cooldownMs);
+  const providerId = resolveProviderId(provider);
+  const codeBuddyQuotaExhausted = ["codebuddy-cn", "codebuddy-intl"].includes(providerId)
+    && isQuotaExhaustionError(errorText);
+  const lockModel = githubResetAtMs || codeBuddyQuotaExhausted ? null : model;
+  const lockUpdate = buildModelLockUpdate(lockModel, cooldownMs);
 
   const deprioUpdate = deprioitizeUntilMs ? { [DEPRIORITIZE_FIELD]: deprioitizeUntilMs } : {};
 
   await updateProviderConnection(connectionId, {
     ...lockUpdate,
     ...deprioUpdate,
+    ...(codeBuddyQuotaExhausted ? { isActive: false } : {}),
     testStatus: "unavailable",
     lastError: reason,
     errorCode: status,
@@ -413,7 +425,7 @@ export async function markAccountUnavailable(connectionId, status, errorText, pr
   });
   invalidateProviderAccountPool(provider);
 
-  await recordLock(connectionId, githubResetAtMs ? null : model, {
+  await recordLock(connectionId, lockModel, {
     tier, reason, expiresAt: new Date(Date.now() + cooldownMs).toISOString(),
   }).catch(() => {});
 

package/src/translator/formats/responsesApi.js

@@ -1,11 +1,11 @@
 import { ROLE, OPENAI_BLOCK, RESPONSES_ITEM } from "../schema/index.js";
 
 export function normalizeResponsesInput(input) {
-  if (typeof input === "string") {
-    return [{ type: RESPONSES_ITEM.MESSAGE, role: ROLE.USER, content: [{ type: RESPONSES_ITEM.INPUT_TEXT, text: input }] }];
-  }
-  if (Array.isArray(input)) {
-    return input;
+  if (typeof input === "string") {
+    return [{ type: RESPONSES_ITEM.MESSAGE, role: ROLE.USER, content: [{ type: RESPONSES_ITEM.INPUT_TEXT, text: input }] }];
+  }
+  if (Array.isArray(input)) {
+    return input;
   }
   return null;
 }
@@ -109,7 +109,9 @@ export function convertResponsesApiFormat(body) {
   delete result.input;
   delete result.instructions;
   delete result.include;
-  delete result.prompt_cache_key;
+  // Preserve a client-supplied prompt_cache_key (explicit conversation identity for
+  // upstream prompt caching); drop only when absent/empty.
+  if (!result.prompt_cache_key) delete result.prompt_cache_key;
   delete result.store;
   delete result.reasoning;
 

package/src/translator/request/openai-responses.js

@@ -209,7 +209,10 @@ export function openaiResponsesToOpenAIRequest(model, body, stream, credentials)
   delete result.input;
   delete result.instructions;
   delete result.include;
-  delete result.prompt_cache_key;
+  // Preserve a client-supplied prompt_cache_key: it is the explicit conversation
+  // identity upstream caches on, so dropping it would guarantee cache misses.
+  // (An empty/absent value is dropped — never send an empty param upstream.)
+  if (!result.prompt_cache_key) delete result.prompt_cache_key;
   delete result.store;
   if (typeof result.reasoning?.effort === "string") {
     result.reasoning_effort = result.reasoning.effort;
@@ -269,18 +272,18 @@ export function openaiToOpenAIResponsesRequest(model, body, stream, credentials)
     store: false
   };
 
-  const systemParts = [];
+  const systemParts = [];
   const messages = body.messages || [];
 
   for (const msg of messages) {
-    if (msg.role === ROLE.SYSTEM || msg.role === ROLE.DEVELOPER) {
-      const content = typeof msg.content === "string"
-        ? msg.content
-        : Array.isArray(msg.content)
-          ? msg.content.map((part) => part?.text || "").filter(Boolean).join("\n")
-          : "";
-      if (content) systemParts.push(content);
-      continue;
+    if (msg.role === ROLE.SYSTEM || msg.role === ROLE.DEVELOPER) {
+      const content = typeof msg.content === "string"
+        ? msg.content
+        : Array.isArray(msg.content)
+          ? msg.content.map((part) => part?.text || "").filter(Boolean).join("\n")
+          : "";
+      if (content) systemParts.push(content);
+      continue;
     }
 
     if (msg.role === ROLE.USER || msg.role === ROLE.ASSISTANT) {
@@ -342,7 +345,7 @@ export function openaiToOpenAIResponsesRequest(model, body, stream, credentials)
     }
   }
 
-  if (systemParts.length > 0) result.instructions = systemParts.join("\n\n");
+  if (systemParts.length > 0) result.instructions = systemParts.join("\n\n");
 
   if (body.tools && Array.isArray(body.tools)) {
     result.tools = body.tools.map(tool => {
@@ -360,10 +363,10 @@ export function openaiToOpenAIResponsesRequest(model, body, stream, credentials)
   }
 
   if (body.temperature !== undefined) result.temperature = body.temperature;
-  // The Responses API uses `max_output_tokens`; sending the Chat Completions
-  // field `max_tokens` makes OpenCode reject Muse Spark requests with 400.
-  if (body.max_output_tokens !== undefined) result.max_output_tokens = body.max_output_tokens;
-  else if (body.max_tokens !== undefined) result.max_output_tokens = body.max_tokens;
+  // The Responses API uses `max_output_tokens`; sending the Chat Completions
+  // field `max_tokens` makes OpenCode reject Muse Spark requests with 400.
+  if (body.max_output_tokens !== undefined) result.max_output_tokens = body.max_output_tokens;
+  else if (body.max_tokens !== undefined) result.max_output_tokens = body.max_tokens;
   if (body.top_p !== undefined) result.top_p = body.top_p;
   if (body.reasoning !== undefined) result.reasoning = body.reasoning;
   if (body.reasoning_effort !== undefined) result.reasoning = { effort: body.reasoning_effort, summary: "auto" };

package/src/utils/error.js

@@ -45,7 +45,7 @@ export async function parseUpstreamError(response, executor = null) {
       if (parsed && typeof parsed === "object") {
         return {
           statusCode: parsed.status || response.status,
-          message: executor?.provider === "codebuddy-intl" && parsed.message
+          message: ["codebuddy-cn", "codebuddy-intl"].includes(executor?.provider) && parsed.message
             ? parsed.message
             : DEFAULT_ERROR_MESSAGES[response.status] || "Upstream provider request failed",
           resetsAtMs: parsed.resetsAtMs,

package/src/utils/sessionManager.js

@@ -90,7 +90,15 @@ function extractAntigravitySession(body) {
     return m ? normalizeSessionId(m[1]) : null;
 }
 
-function extractClientSessionId(headers, body, scope = "") {
+/**
+ * @param {Record<string, string>|null|undefined} headers
+ * @param {unknown} body
+ * @param {string} [scope]
+ * @param {{ includeRequestId?: boolean }} [opts]
+ * @returns {string|null}
+ */
+export function extractClientSessionId(headers, body, scope = "", opts = {}) {
+    const { includeRequestId = true } = opts;
     const claude = extractClaudeCodeSession(body?.metadata?.user_id);
     if (claude) return `claude:${claude}`;
     const antigravity = extractAntigravitySession(body);
@@ -99,7 +107,7 @@ function extractClientSessionId(headers, body, scope = "") {
         const v = headerValue(headers, key);
         if (v) return v;
     }
-    const requestId = scope === "kiro" ? null : headerValue(headers, "x-client-request-id");
+    const requestId = (scope === "kiro" || !includeRequestId) ? null : headerValue(headers, "x-client-request-id");
     if (requestId) return requestId;
     const fromBody =
         normalizeSessionId(body?.prompt_cache_key) ||