vg-coder-cli 2.0.61 → 2.0.63

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vg-coder-cli",
-  "version": "2.0.61",
+  "version": "2.0.63",
   "description": "🚀 CLI tool to analyze projects, concatenate source files, count tokens, and export HTML with syntax highlighting and copy functionality",
   "main": "src/index.js",
   "bin": {

package/src/server/api-server.js

@@ -668,6 +668,25 @@ class ApiServer {
         res.json({ workers: taskQueue.listWorkers() });
     });
 
+    // Reset cooldown manually. Useful khi operator chắc chắn AI Studio đã hết
+    // rate-limit (vd quan sát qua noVNC submit prompt được) nhưng server timer
+    // 30 phút chưa expire. Body: { email: "alice@gmail.com" } hoặc {} cho tất cả.
+    this.app.post('/api/workers/reset-cooldown', (req, res) => {
+        const targetEmail = (req.body?.email || '').toString().toLowerCase().trim();
+        const reset = [];
+        for (const w of taskQueue.workers.values()) {
+            if (w.status !== 'rate_limited') continue;
+            if (targetEmail && w.email !== targetEmail) continue;
+            w.status = 'idle';
+            w.cooldownUntil = 0;
+            w.lastFailureCode = null;
+            w.lastFailureMessage = null;
+            reset.push({ id: w.id, email: w.email });
+        }
+        if (reset.length) setImmediate(() => taskQueue._drain());
+        res.json({ ok: true, reset, count: reset.length });
+    });
+
     this.app.get('/api/launchers', (req, res) => {
         res.json({ launchers: taskQueue.listLaunchers() });
     });

package/src/server/task-queue.js

@@ -109,13 +109,21 @@ class TaskQueue {
 
   listWorkers() {
     const out = [];
+    const now = Date.now();
     for (const w of this.workers.values()) {
+      const cd = w.cooldownUntil || 0;
+      const remainingMs = cd > now ? cd - now : 0;
       out.push({
         id: w.id,
         email: w.email,
         status: w.status,
         currentTaskId: w.currentTaskId,
-        cooldownUntil: w.cooldownUntil || 0,
+        cooldownUntil: cd,
+        cooldownUntilISO: cd ? new Date(cd).toISOString() : null,
+        cooldownRemainingMs: remainingMs,
+        cooldownRemainingMinutes: Math.ceil(remainingMs / 60_000),
+        lastFailureCode: w.lastFailureCode || null,
+        lastFailureMessage: w.lastFailureMessage || null,
         lastSeen: w.lastSeen,
         meta: { domain: w.meta?.domain, chatId: w.meta?.chatId, registeredAt: w.meta?.registeredAt }
       });
@@ -616,10 +624,14 @@ class TaskQueue {
     if (RL_CODES.has(code)) {
       worker.status = 'rate_limited';
       worker.cooldownUntil = Date.now() + RATE_LIMIT_COOLDOWN_MS;
+      worker.lastFailureCode = code;
+      worker.lastFailureMessage = message.slice(0, 300);
       worker.currentTaskId = null;
       console.log(chalk.yellow(`[TaskQueue] Worker ${worker.email} rate-limited until ${new Date(worker.cooldownUntil).toLocaleTimeString()}`));
     } else {
       // Plain failure — worker becomes idle again
+      worker.lastFailureCode = code;
+      worker.lastFailureMessage = message.slice(0, 300);
       worker.currentTaskId = null;
       worker.status = 'idle';
     }

package/src/server/views/js/features/task-worker.js

@@ -133,12 +133,15 @@ async function fetchAsFile(url, name, mime) {
 async function waitForNewAssistantTurn(baselineCount, timeoutMs = 90_000) {
     const start = Date.now();
     while (Date.now() - start < timeoutMs) {
-        // Surface rate-limit / quota errors early — no point waiting for a turn
-        // that will never appear.
-        const rl = detectRateLimit();
-        if (rl) { const e = new Error(rl.message); e.code = rl.code; throw e; }
-
+        // Chỉ surface rate-limit/quota khi turn count đã vượt baseline. Trước
+        // đó turn cuối vẫn là turn cũ (có thể là error từ session restore) —
+        // detectRateLimit match phải khớp với turn MỚI.
         const turns = document.querySelectorAll('ms-chat-turn');
+        if (turns.length > baselineCount) {
+            // Turn mới đã render — bây giờ mới check (toast vẫn check global ok)
+            const rl = detectRateLimit();
+            if (rl) { const e = new Error(rl.message); e.code = rl.code; throw e; }
+        }
         if (turns.length >= baselineCount + 2) {
             const last = turns[turns.length - 1];
             const role = last.querySelector('[data-turn-role]')?.getAttribute('data-turn-role');
@@ -164,8 +167,15 @@ function detectRateLimit() {
         if (/failed to generate/i.test(t))   return { code: 'generation_failed', message: t.slice(0, 500) };
     }
 
-    // Inline model-error inside the last assistant turn
-    const modelErr = document.querySelector('.model-error');
+    // Inline model-error — CHỈ check trong turn CUỐI (response cho prompt vừa
+    // submit). AI Studio restore chat session từ cookies → turn cũ với
+    // .model-error rate-limit có thể vẫn trong DOM dù limit đã hết.
+    // Global querySelector('.model-error') match turn cũ → mark cooldown
+    // infinite loop. Scope theo last ms-chat-turn để chỉ trigger trên response
+    // mới.
+    const turns = document.querySelectorAll('ms-chat-turn');
+    const lastTurn = turns[turns.length - 1];
+    const modelErr = lastTurn?.querySelector('.model-error');
     if (modelErr) {
         const text = (modelErr.textContent || '').trim();
         if (/rate limit/i.test(text))       return { code: 'rate_limit_exceeded', message: text };