@protolabsai/proto 0.25.13 → 0.25.15

package/cli.js

@@ -170894,7 +170894,7 @@ __export(geminiContentGenerator_exports, {
   createGeminiContentGenerator: () => createGeminiContentGenerator
 });
 function createGeminiContentGenerator(config2, gcConfig) {
-  const version2 = "0.25.13";
+  const version2 = "0.25.15";
   const userAgent2 = config2.userAgent || `QwenCode/${version2} (${process.platform}; ${process.arch})`;
   const baseHeaders = {
     "User-Agent": userAgent2
@@ -171950,13 +171950,78 @@ function extractCuratedHistory2(comprehensiveHistory) {
   }
   return curatedHistory;
 }
+function hasTruncationCascade(contents) {
+  if (contents.length === 0)
+    return false;
+  const last2 = contents[contents.length - 1];
+  if (last2.role !== "user")
+    return false;
+  return last2.parts?.some((p2) => {
+    const err2 = p2.functionResponse?.response?.["error"];
+    return typeof err2 === "string" && err2.includes(TRUNCATION_CASCADE_MARKER);
+  }) ?? false;
+}
+function trimLargeToolResponsesFromContext(contents) {
+  return contents.map((content) => {
+    if (content.role !== "user" || !content.parts)
+      return content;
+    const needsTrim = content.parts.some((p2) => {
+      const out2 = p2.functionResponse?.response?.["output"];
+      return typeof out2 === "string" && out2.length > LARGE_TOOL_RESPONSE_TRIM_CHARS;
+    });
+    if (!needsTrim)
+      return content;
+    return {
+      ...content,
+      parts: content.parts.map((p2) => {
+        const out2 = p2.functionResponse?.response?.["output"];
+        if (typeof out2 !== "string" || out2.length <= LARGE_TOOL_RESPONSE_TRIM_CHARS)
+          return p2;
+        const trimmed2 = out2.slice(0, LARGE_TOOL_RESPONSE_TRIM_CHARS) + `
+
+[... output trimmed from ${out2.length} to ${LARGE_TOOL_RESPONSE_TRIM_CHARS} chars to reduce context size ...]`;
+        return {
+          ...p2,
+          functionResponse: {
+            ...p2.functionResponse,
+            response: {
+              ...p2.functionResponse.response,
+              output: trimmed2
+            }
+          }
+        };
+      })
+    };
+  });
+}
+function trimToolErrorsFromContext(contents, maxTrimPairs = 6) {
+  const trimmed2 = [...contents];
+  let trimmed_count = 0;
+  while (trimmed2.length >= 2 && trimmed_count < maxTrimPairs) {
+    const last2 = trimmed2[trimmed2.length - 1];
+    const prev = trimmed2[trimmed2.length - 2];
+    if (last2.role !== "user")
+      break;
+    const allErrors = last2.parts?.every((p2) => p2.functionResponse !== void 0 && typeof p2.functionResponse.response?.["error"] === "string");
+    if (!allErrors)
+      break;
+    if (prev.role !== "model")
+      break;
+    const hasToolCall2 = prev.parts?.some((p2) => p2.functionCall !== void 0);
+    if (!hasToolCall2)
+      break;
+    trimmed2.splice(trimmed2.length - 2, 2);
+    trimmed_count++;
+  }
+  return trimmed2;
+}
 function isSchemaDepthError(errorMessage) {
   return errorMessage.includes("maximum schema depth exceeded");
 }
 function isInvalidArgumentError(errorMessage) {
   return errorMessage.includes("Request contains an invalid argument");
 }
-var debugLogger22, StreamEventType, INVALID_CONTENT_RETRY_OPTIONS, INVALID_STREAM_RETRY_CONFIG, RATE_LIMIT_RETRY_OPTIONS, InvalidStreamError, GeminiChat;
+var debugLogger22, StreamEventType, INVALID_CONTENT_RETRY_OPTIONS, INVALID_STREAM_RETRY_CONFIG, RATE_LIMIT_RETRY_OPTIONS, TRUNCATION_CASCADE_MARKER, LARGE_TOOL_RESPONSE_TRIM_CHARS, InvalidStreamError, GeminiChat;
 var init_geminiChat = __esm({
   "packages/core/dist/src/core/geminiChat.js"() {
     "use strict";
@@ -171995,6 +172060,11 @@ var init_geminiChat = __esm({
     __name(isValidContentPart, "isValidContentPart");
     __name(validateHistory2, "validateHistory");
     __name(extractCuratedHistory2, "extractCuratedHistory");
+    TRUNCATION_CASCADE_MARKER = "truncated due to max_tokens limit";
+    LARGE_TOOL_RESPONSE_TRIM_CHARS = 1e4;
+    __name(hasTruncationCascade, "hasTruncationCascade");
+    __name(trimLargeToolResponsesFromContext, "trimLargeToolResponsesFromContext");
+    __name(trimToolErrorsFromContext, "trimToolErrorsFromContext");
     InvalidStreamError = class extends Error {
       static {
         __name(this, "InvalidStreamError");
@@ -172072,7 +172142,12 @@ var init_geminiChat = __esm({
         this.sendPromise = streamDonePromise;
         const userContent = createUserContent(params.message);
         this.history.push(userContent);
-        const requestContents = this.getHistory(true);
+        let requestContents = this.getHistory(true);
+        if (hasTruncationCascade(requestContents)) {
+          const withoutErrors = trimToolErrorsFromContext(requestContents);
+          requestContents = trimLargeToolResponsesFromContext(withoutErrors);
+          debugLogger22.warn(`MAX_TOKENS cascade detected: trimmed context from ${this.getHistory(true).length} to ${requestContents.length} entries and capped large tool responses.`);
+        }
         const self2 = this;
         return async function* () {
           try {
@@ -172125,6 +172200,21 @@ var init_geminiChat = __esm({
                   await new Promise((res) => setTimeout(res, delayMs));
                   continue;
                 }
+                if (isTransientStreamError && error40.type === "NO_RESPONSE_TEXT") {
+                  const trimmedContents = trimToolErrorsFromContext(requestContents);
+                  if (trimmedContents.length < requestContents.length) {
+                    debugLogger22.warn(`NO_RESPONSE_TEXT: retrying with trimmed context (removed ${requestContents.length - trimmedContents.length} tool-error messages)`);
+                    try {
+                      const stream2 = await self2.makeApiCallAndProcessStream(model, trimmedContents, params, prompt_id);
+                      for await (const chunk of stream2) {
+                        yield { type: StreamEventType.CHUNK, value: chunk };
+                      }
+                      lastError = null;
+                    } catch {
+                    }
+                  }
+                  break;
+                }
                 if (isTransientStreamError) {
                   break;
                 }
@@ -414871,7 +414961,7 @@ __name(getPackageJson, "getPackageJson");
 // packages/cli/src/utils/version.ts
 async function getCliVersion() {
   const pkgJson = await getPackageJson();
-  return "0.25.13";
+  return "0.25.15";
 }
 __name(getCliVersion, "getCliVersion");
 
@@ -422643,7 +422733,7 @@ var formatDuration = /* @__PURE__ */ __name((milliseconds) => {
 
 // packages/cli/src/generated/git-commit.ts
 init_esbuild_shims();
-var GIT_COMMIT_INFO = "05b963410";
+var GIT_COMMIT_INFO = "eda760681";
 
 // packages/cli/src/utils/systemInfo.ts
 async function getNpmVersion() {
@@ -490122,7 +490212,7 @@ var QwenAgent = class {
   async initialize(args2) {
     this.clientCapabilities = args2.clientCapabilities;
     const authMethods = buildAuthMethods();
-    const version2 = "0.25.13";
+    const version2 = "0.25.15";
     return {
       protocolVersion: PROTOCOL_VERSION,
       agentInfo: {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@protolabsai/proto",
-  "version": "0.25.13",
+  "version": "0.25.15",
   "description": "proto - AI-powered coding agent",
   "repository": {
     "type": "git",
@@ -21,7 +21,7 @@
     "bundled"
   ],
   "config": {
-    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.25.13"
+    "sandboxImageUri": "ghcr.io/qwenlm/qwen-code:0.25.15"
   },
   "dependencies": {},
   "optionalDependencies": {