cc-claw 0.13.0 → 0.13.2

package/dist/cli.js

@@ -72,7 +72,7 @@ var VERSION;
 var init_version = __esm({
   "src/version.ts"() {
     "use strict";
-    VERSION = true ? "0.13.0" : (() => {
+    VERSION = true ? "0.13.2" : (() => {
       try {
         return JSON.parse(readFileSync(join2(process.cwd(), "package.json"), "utf-8")).version ?? "unknown";
       } catch {
@@ -9684,6 +9684,9 @@ function stopAgent(chatId) {
   }
   return true;
 }
+function getGeminiFallback(model2) {
+  return GEMINI_FALLBACK_CHAIN[model2] ?? null;
+}
 function spawnQuery(adapter, config2, model2, cancelState, thinkingLevel, timeoutMs, maxTurns, opts) {
   const effectiveTimeout = timeoutMs ?? SPAWN_TIMEOUT_MS;
   return new Promise((resolve, reject) => {
@@ -9727,10 +9730,11 @@ function spawnQuery(adapter, config2, model2, cancelState, thinkingLevel, timeou
     let firstLine = true;
     const frTimeoutMs = opts?.firstResponseTimeoutMs ?? FIRST_RESPONSE_TIMEOUT_MS;
     let firstResponseTimer;
+    let gotModelContent = false;
     if (frTimeoutMs > 0) {
       firstResponseTimer = setTimeout(() => {
-        if (firstLine) {
-          warn(`[agent] First-response timeout after ${frTimeoutMs}ms for ${adapter.id} \u2014 no NDJSON output received, killing`);
+        if (!gotModelContent) {
+          warn(`[agent] First-response timeout after ${frTimeoutMs}ms for ${adapter.id} \u2014 no model content received (init may have arrived), killing`);
           killProcessGroup(proc, "SIGTERM");
           timedOut = true;
           cancelState.__firstResponseTimeout = true;
@@ -9743,10 +9747,6 @@ function spawnQuery(adapter, config2, model2, cancelState, thinkingLevel, timeou
       if (firstLine) {
         log(`[agent] First CLI message after ${elapsed()}`);
         firstLine = false;
-        if (firstResponseTimer) {
-          clearTimeout(firstResponseTimer);
-          firstResponseTimer = void 0;
-        }
       }
       let msg;
       try {
@@ -9766,12 +9766,26 @@ function spawnQuery(adapter, config2, model2, cancelState, thinkingLevel, timeou
             if (ev.sessionId) sessionId = ev.sessionId;
             break;
           case "text":
+            if (!gotModelContent) {
+              gotModelContent = true;
+              if (firstResponseTimer) {
+                clearTimeout(firstResponseTimer);
+                firstResponseTimer = void 0;
+              }
+            }
             if (ev.text) {
               accumulatedText = appendTextChunk(accumulatedText, ev.text);
               if (opts?.onStream) opts.onStream(ev.text);
             }
             break;
           case "tool_start":
+            if (!gotModelContent) {
+              gotModelContent = true;
+              if (firstResponseTimer) {
+                clearTimeout(firstResponseTimer);
+                firstResponseTimer = void 0;
+              }
+            }
             sawToolEvents = true;
             if (opts?.onToolAction && ev.toolName) {
               const toolInput = ev.toolInput ?? {};
@@ -9821,6 +9835,13 @@ function spawnQuery(adapter, config2, model2, cancelState, thinkingLevel, timeou
             }
             break;
           case "result":
+            if (!gotModelContent) {
+              gotModelContent = true;
+              if (firstResponseTimer) {
+                clearTimeout(firstResponseTimer);
+                firstResponseTimer = void 0;
+              }
+            }
             sawResultEvent = true;
             log(`[agent] Result received at ${elapsed()}`);
             resultText = ev.resultText || accumulatedText;
@@ -9934,12 +9955,14 @@ async function spawnGeminiWithRotation(chatId, adapter, baseConfig, configWithSe
       const isTimeout = errMsg.startsWith(FIRST_RESPONSE_TIMEOUT_ERROR);
       const isExhausted = /RESOURCE.?EXHAUSTED|resource has been exhausted/i.test(errMsg);
       if ((isTimeout || isExhausted) && GEMINI_DOWNGRADE_MODELS.has(model2)) {
+        const fallbackModel = getGeminiFallback(model2);
         const reason = isTimeout ? "No response within timeout" : "Resource exhausted";
-        warn(`[agent:gemini-rotation] ${reason} on ${model2} (${slotLabel}) \u2014 downgrading to ${GEMINI_FALLBACK_MODEL}`);
-        onModelDowngrade?.(model2, GEMINI_FALLBACK_MODEL, `${reason} \u2014 model likely overloaded`);
+        warn(`[agent:gemini-rotation] ${reason} on ${model2} (${slotLabel}) \u2014 downgrading to ${fallbackModel}`);
+        onModelDowngrade?.(model2, fallbackModel, `${reason} \u2014 model likely overloaded`);
+        setModel(chatId, fallbackModel);
         const fallbackConfig = adapter.buildSpawnConfig({
           prompt: effectiveConfig.args[effectiveConfig.args.indexOf("-p") + 1],
-          model: GEMINI_FALLBACK_MODEL,
+          model: fallbackModel,
           permMode: "yolo",
           allowedTools: [],
           cwd: effectiveConfig.cwd,
@@ -9956,7 +9979,9 @@ async function spawnGeminiWithRotation(chatId, adapter, baseConfig, configWithSe
             }
           }
         }
-        return await spawnQuery(adapter, fallbackConfig, GEMINI_FALLBACK_MODEL, cancelState, thinkingLevel, timeoutMs, maxTurns, { ...opts, envOverride: env, firstResponseTimeoutMs: 0 });
+        const fallbackResult = await spawnQuery(adapter, fallbackConfig, fallbackModel, cancelState, thinkingLevel, timeoutMs, maxTurns, { ...opts, envOverride: env });
+        fallbackResult.resolvedModel = fallbackModel;
+        return fallbackResult;
       }
       const quotaClass = classifyGeminiQuota(errMsg);
       if (quotaClass === "rate_limited") {
@@ -10068,12 +10093,14 @@ async function askAgentImpl(chatId, userMessage, opts) {
         const isTimeout = errMsg.startsWith(FIRST_RESPONSE_TIMEOUT_ERROR);
         const isExhausted = /RESOURCE.?EXHAUSTED|resource has been exhausted/i.test(errMsg);
         if (adapter.id === "gemini" && (isTimeout || isExhausted) && GEMINI_DOWNGRADE_MODELS.has(resolvedModel)) {
+          const fallbackModel = getGeminiFallback(resolvedModel);
           const reason = isTimeout ? "No response within timeout" : "Resource exhausted";
-          warn(`[agent] ${reason} on ${resolvedModel} (no rotation) \u2014 downgrading to ${GEMINI_FALLBACK_MODEL}`);
-          onModelDowngrade?.(chatId, resolvedModel, GEMINI_FALLBACK_MODEL, `${reason} \u2014 model likely overloaded`);
+          warn(`[agent] ${reason} on ${resolvedModel} (no rotation) \u2014 downgrading to ${fallbackModel}`);
+          onModelDowngrade?.(chatId, resolvedModel, fallbackModel, `${reason} \u2014 model likely overloaded`);
+          setModel(chatId, fallbackModel);
           const fallbackConfig = adapter.buildSpawnConfig({
             prompt: configWithSession.args[configWithSession.args.indexOf("-p") + 1],
-            model: GEMINI_FALLBACK_MODEL,
+            model: fallbackModel,
             permMode: mode,
             allowedTools,
             cwd: resolvedCwd,
@@ -10083,7 +10110,8 @@ async function askAgentImpl(chatId, userMessage, opts) {
           if (mcpConfigPath) {
             fallbackConfig.args = injectMcpConfig(adapter.id, fallbackConfig.args, mcpConfigPath);
           }
-          result = await spawnQuery(adapter, fallbackConfig, GEMINI_FALLBACK_MODEL, cancelState, thinkingLevel, timeoutMs, maxTurns, { ...spawnOpts, firstResponseTimeoutMs: 0 });
+          result = await spawnQuery(adapter, fallbackConfig, fallbackModel, cancelState, thinkingLevel, timeoutMs, maxTurns, spawnOpts);
+          result.resolvedModel = fallbackModel;
         } else {
           throw err;
         }
@@ -10154,7 +10182,8 @@ async function askAgentImpl(chatId, userMessage, opts) {
   return {
     text: result.resultText || `(No response from ${adapter.displayName})`,
     sessionId: result.sessionId,
-    usage: { input: result.input, output: result.output, cacheRead: result.cacheRead, contextSize: result.contextSize }
+    usage: { input: result.input, output: result.output, cacheRead: result.cacheRead, contextSize: result.contextSize },
+    resolvedModel: result.resolvedModel
   };
 }
 function getMcpConfigPath(chatId) {
@@ -10168,7 +10197,7 @@ function injectMcpConfig(adapterId, args, mcpConfigPath) {
   if (!flag) return args;
   return [...args, ...flag, mcpConfigPath];
 }
-var activeChats, chatLocks, SPAWN_TIMEOUT_MS, FIRST_RESPONSE_TIMEOUT_MS, FIRST_RESPONSE_TIMEOUT_ERROR, GEMINI_FALLBACK_MODEL, GEMINI_DOWNGRADE_MODELS, MCP_CONFIG_FLAG;
+var activeChats, chatLocks, SPAWN_TIMEOUT_MS, FIRST_RESPONSE_TIMEOUT_MS, FIRST_RESPONSE_TIMEOUT_ERROR, GEMINI_FALLBACK_CHAIN, GEMINI_DOWNGRADE_MODELS, MCP_CONFIG_FLAG;
 var init_agent = __esm({
   "src/agent.ts"() {
     "use strict";
@@ -10191,10 +10220,13 @@ var init_agent = __esm({
     activeChats = /* @__PURE__ */ new Map();
     chatLocks = /* @__PURE__ */ new Map();
     SPAWN_TIMEOUT_MS = 10 * 60 * 1e3;
-    FIRST_RESPONSE_TIMEOUT_MS = parseInt(process.env.GEMINI_FIRST_RESPONSE_TIMEOUT_MS ?? "15000", 10);
+    FIRST_RESPONSE_TIMEOUT_MS = parseInt(process.env.GEMINI_FIRST_RESPONSE_TIMEOUT_MS ?? "30000", 10);
     FIRST_RESPONSE_TIMEOUT_ERROR = "FIRST_RESPONSE_TIMEOUT";
-    GEMINI_FALLBACK_MODEL = "gemini-2.5-pro";
-    GEMINI_DOWNGRADE_MODELS = /* @__PURE__ */ new Set(["gemini-3.1-pro-preview"]);
+    GEMINI_FALLBACK_CHAIN = {
+      "gemini-3.1-pro-preview": "gemini-2.5-pro",
+      "gemini-2.5-pro": "gemini-3.1-flash-preview"
+    };
+    GEMINI_DOWNGRADE_MODELS = new Set(Object.keys(GEMINI_FALLBACK_CHAIN));
     MCP_CONFIG_FLAG = {
       claude: ["--mcp-config"]
     };
@@ -15202,12 +15234,20 @@ You're still in discussion mode \u2014 try again or click a button to exit.`, {
     const sigEnabled = getModelSignature(chatId);
     if (sigEnabled === "on" && responseText && !responseText.startsWith("(No response")) {
       const adapter = getAdapterForChat(chatId);
-      const modelId = model2 ?? adapter.defaultModel;
+      const modelId = response.resolvedModel ?? model2 ?? adapter.defaultModel;
       const thinking2 = getThinkingLevel(chatId) || "auto";
       const shortModel = formatModelShort(modelId);
+      let slotTag = "";
+      if (adapter.id === "gemini") {
+        const slotId = getChatGeminiSlotId(chatId);
+        if (slotId) {
+          const slot = getGeminiSlots().find((s) => s.id === slotId);
+          if (slot) slotTag = ` \xB7 ${slot.label || `slot-${slot.id}`}`;
+        }
+      }
       responseText += `
 
-\u{1F9E0} [${shortModel} | ${capitalize(thinking2)}] \u23F1\uFE0F ${elapsedSec}s`;
+\u{1F9E0} [${shortModel} | ${capitalize(thinking2)}${slotTag}] \u23F1\uFE0F ${elapsedSec}s`;
     }
     if (observedSubagents.size > 0) {
       const names = [...observedSubagents].join(", ");
@@ -16474,7 +16514,11 @@ Result: ${task.result.slice(0, 500)}` : ""
         pinChatGeminiSlot(chatId, slotId);
         const label2 = slot.label || `slot-${slot.id}`;
         const icon = slot.slotType === "oauth" ? "\u{1F468}\u{1F3FD}\u200D\u{1F4BB}" : "\u{1F511}";
-        await channel.sendText(chatId, `Pinned to ${icon} <b>${label2}</b>`, { parseMode: "html" });
+        const rotationMode = getGeminiRotationMode();
+        const rotationLabels = { off: "off", all: "all slots", accounts: "accounts only", keys: "keys only" };
+        const rotationNote = rotationMode === "off" ? "Rotation is off \u2014 only this account will be used." : `Rotation: ${rotationLabels[rotationMode]} (this is the starting slot).`;
+        await channel.sendText(chatId, `${icon} Account set to <b>${label2}</b>
+${rotationNote}`, { parseMode: "html" });
       }
     }
     return;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cc-claw",
-  "version": "0.13.0",
+  "version": "0.13.2",
   "description": "CC-Claw: Personal AI assistant on Telegram — multi-backend (Claude, Gemini, Codex, Cursor), sub-agent orchestration, MCP management",
   "type": "module",
   "main": "dist/cli.js",