@protolabsai/proto 0.55.3 → 0.55.4

package/cli.js

@@ -146451,7 +146451,7 @@ var init_constants2 = __esm({
   "packages/core/dist/src/core/openaiContentGenerator/constants.js"() {
     "use strict";
     init_esbuild_shims();
-    DEFAULT_TIMEOUT = 12e4;
+    DEFAULT_TIMEOUT = 6e5;
     DEFAULT_MAX_RETRIES = 3;
     DEFAULT_OPENAI_BASE_URL = "https://api.openai.com/v1";
     DEFAULT_DASHSCOPE_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1";
@@ -155562,7 +155562,8 @@ var init_pipeline = __esm({
         raiseAbortListenerCap(request3.config?.abortSignal);
         return this.executeWithErrorHandling(request3, userPromptId, true, effectiveModel, async (openaiRequest, context2) => {
           const stream2 = await this.client.chat.completions.create(openaiRequest, {
-            signal: request3.config?.abortSignal
+            signal: request3.config?.abortSignal,
+            timeout: 0
           });
           return this.processStreamWithLogging(stream2, context2, request3);
         });
@@ -167976,7 +167977,11 @@ var init_anthropicContentGenerator = __esm({
             stream: true
           };
           const stream2 = await this.client.messages.create(streamingRequest, {
-            signal: request3.config?.abortSignal
+            signal: request3.config?.abortSignal,
+            // Disable the client-level timeout for streaming requests. Same
+            // rationale as the OpenAI pipeline: long agent turns can exceed
+            // any fixed wall-clock timeout.
+            timeout: 0
           });
           return this.processStreamWithSpan(stream2, span, startTime, modelId, logPrompts);
         } catch (error40) {
@@ -168684,7 +168689,7 @@ __export(geminiContentGenerator_exports, {
   createGeminiContentGenerator: () => createGeminiContentGenerator
 });
 function createGeminiContentGenerator(config2, gcConfig) {
-  const version2 = "0.55.3";
+  const version2 = "0.55.4";
   const userAgent2 = config2.userAgent || `QwenCode/${version2} (${process.platform}; ${process.arch})`;
   const baseHeaders = {
     "User-Agent": userAgent2
@@ -169198,7 +169203,7 @@ var init_turn = __esm({
     init_errors();
     init_thoughtUtils();
     init_streamStall();
-    STREAM_STALL_TIMEOUT_MS = parseInt(process.env["PROTO_STREAM_STALL_TIMEOUT_MS"] ?? "300000", 10);
+    STREAM_STALL_TIMEOUT_MS = parseInt(process.env["PROTO_STREAM_STALL_TIMEOUT_MS"] ?? "660000", 10);
     (function(GeminiEventType2) {
       GeminiEventType2["Content"] = "content";
       GeminiEventType2["ToolCallRequest"] = "tool_call_request";
@@ -169319,16 +169324,7 @@ ${[...this.pendingCitations].sort().join("\n")}`
             return;
           }
           if (e4 instanceof StreamStallError) {
-            yield {
-              type: GeminiEventType.Error,
-              value: {
-                error: {
-                  message: e4.message,
-                  status: void 0
-                }
-              }
-            };
-            return;
+            throw e4;
           }
           const error40 = toFriendlyError(e4);
           if (error40 instanceof UnauthorizedError) {
@@ -172580,6 +172576,13 @@ function trimToolErrorsFromContext(contents, maxTrimPairs = 6) {
   }
   return trimmed2;
 }
+function isSdkTransportTimeoutError(error40) {
+  if (!(error40 instanceof Error))
+    return false;
+  const msg = error40.message.toLowerCase();
+  return msg.includes("request timed out") || msg.includes("request timeout") || msg.includes("connection timeout") || msg.includes("timed out") || error40.name === "TimeoutError" || // OpenAI SDK wraps transport timeouts with these codes
+  error40.code === "ETIMEDOUT" || error40.code === "ESOCKETTIMEDOUT";
+}
 function isSchemaDepthError(errorMessage) {
   return errorMessage.includes("maximum schema depth exceeded");
 }
@@ -172601,6 +172604,7 @@ var init_geminiChat = __esm({
     init_tools();
     init_chatCompressionService();
     init_turn();
+    init_streamStall();
     init_loggers();
     init_chatRecordingService();
     init_types4();
@@ -172644,6 +172648,7 @@ var init_geminiChat = __esm({
         this.type = type;
       }
     };
+    __name(isSdkTransportTimeoutError, "isSdkTransportTimeoutError");
     GeminiChat = class {
       static {
         __name(this, "GeminiChat");
@@ -172758,12 +172763,18 @@ var init_geminiChat = __esm({
                   await new Promise((res) => setTimeout(res, delayMs));
                   continue;
                 }
-                const isTransientStreamError = error40 instanceof InvalidStreamError;
+                const isTransientStreamError = error40 instanceof InvalidStreamError || error40 instanceof StreamStallError || isSdkTransportTimeoutError(error40);
                 if (isTransientStreamError && invalidStreamRetryCount < INVALID_STREAM_RETRY_CONFIG.maxRetries) {
                   invalidStreamRetryCount++;
                   const delayMs = INVALID_STREAM_RETRY_CONFIG.initialDelayMs * invalidStreamRetryCount;
-                  debugLogger25.warn(`Invalid stream [${error40.type}] (retry ${invalidStreamRetryCount}/${INVALID_STREAM_RETRY_CONFIG.maxRetries}). Waiting ${delayMs / 1e3}s before retrying...`);
-                  logContentRetry(self2.config, new ContentRetryEvent(invalidStreamRetryCount - 1, error40.type, delayMs, model));
+                  if (error40 instanceof StreamStallError) {
+                    debugLogger25.warn(`Stream stalled (retry ${invalidStreamRetryCount}/${INVALID_STREAM_RETRY_CONFIG.maxRetries}). Waiting ${delayMs / 1e3}s before retrying...`);
+                  } else if (isSdkTransportTimeoutError(error40)) {
+                    debugLogger25.warn(`SDK transport timeout (retry ${invalidStreamRetryCount}/${INVALID_STREAM_RETRY_CONFIG.maxRetries}). Waiting ${delayMs / 1e3}s before retrying...`);
+                  } else {
+                    debugLogger25.warn(`Invalid stream [${error40.type}] (retry ${invalidStreamRetryCount}/${INVALID_STREAM_RETRY_CONFIG.maxRetries}). Waiting ${delayMs / 1e3}s before retrying...`);
+                  }
+                  logContentRetry(self2.config, new ContentRetryEvent(invalidStreamRetryCount - 1, error40 instanceof StreamStallError ? "STREAM_STALL" : isSdkTransportTimeoutError(error40) ? "SDK_TRANSPORT_TIMEOUT" : error40.type, delayMs, model));
                   yield { type: StreamEventType.RETRY };
                   attempt--;
                   await new Promise((res) => setTimeout(res, delayMs));
@@ -172835,6 +172846,12 @@ var init_geminiChat = __esm({
               if (lastError instanceof InvalidStreamError) {
                 const totalAttempts = invalidStreamRetryCount + 1;
                 logContentRetryFailure(self2.config, new ContentRetryFailureEvent(totalAttempts, lastError.type, model));
+              } else if (lastError instanceof StreamStallError) {
+                const totalAttempts = invalidStreamRetryCount + 1;
+                logContentRetryFailure(self2.config, new ContentRetryFailureEvent(totalAttempts, "STREAM_STALL", model));
+              } else if (isSdkTransportTimeoutError(lastError)) {
+                const totalAttempts = invalidStreamRetryCount + 1;
+                logContentRetryFailure(self2.config, new ContentRetryFailureEvent(totalAttempts, "SDK_TRANSPORT_TIMEOUT", model));
               }
               throw lastError;
             }
@@ -416974,7 +416991,7 @@ __name(getPackageJson, "getPackageJson");
 // packages/cli/src/utils/version.ts
 async function getCliVersion() {
   const pkgJson = await getPackageJson();
-  return "0.55.3";
+  return "0.55.4";
 }
 __name(getCliVersion, "getCliVersion");
 
@@ -425174,7 +425191,7 @@ var formatDuration = /* @__PURE__ */ __name((milliseconds) => {
 
 // packages/cli/src/generated/git-commit.ts
 init_esbuild_shims();
-var GIT_COMMIT_INFO = "72bbb0204";
+var GIT_COMMIT_INFO = "d473ef1d8";
 
 // packages/cli/src/utils/systemInfo.ts
 async function getNpmVersion() {
@@ -494063,7 +494080,7 @@ var QwenAgent = class {
   async initialize(args2) {
     this.clientCapabilities = args2.clientCapabilities;
     const authMethods = buildAuthMethods();
-    const version2 = "0.55.3";
+    const version2 = "0.55.4";
     return {
       protocolVersion: PROTOCOL_VERSION,
       agentInfo: {
@@ -495160,11 +495177,12 @@ main().catch((error40) => {
  *
  * Stream stall detection — per-chunk idle watchdog for async generators.
  *
- * The SDK's overall request timeout (120 s) covers the full lifetime of a
- * request, but undici body/header timeouts are explicitly disabled in
- * runtimeFetchOptions.ts to allow long streaming responses.  This means a
- * stream that *starts* and then *freezes* (connection open, no more chunks
- * arriving) will not be detected until the full 120 s wall clock fires.
+ * The SDK's overall request timeout (configurable via contentGenerator.timeout,
+ * defaults to 600 s) covers the full lifetime of a request, but undici
+ * body/header timeouts are explicitly disabled in runtimeFetchOptions.ts to
+ * allow long streaming responses.  This means a stream that *starts* and then
+ * *freezes* (connection open, no more chunks arriving) will not be detected
+ * until the full SDK wall clock fires.
  *
  * `withChunkTimeout` wraps any AsyncGenerator and races each `.next()` call
  * against a per-chunk idle timer.  If no chunk arrives within `timeoutMs`,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@protolabsai/proto",
-  "version": "0.55.3",
+  "version": "0.55.4",
   "description": "proto - AI-powered coding agent",
   "repository": {
     "type": "git",
@@ -21,7 +21,7 @@
     "bundled"
   ],
   "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.55.3"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.55.4"
   },
   "dependencies": {},
   "optionalDependencies": {