open-agents-ai 0.187.277 → 0.187.280

package/dist/index.js

@@ -264717,12 +264717,12 @@ var init_taskNormalizer = __esm({
 // packages/orchestrator/dist/dispatcher.js
 function estimateComplexity(task) {
   const text = `${task.goal} ${task.constraints.join(" ")} ${task.successCriteria.join(" ")}`;
-  const wordCount = text.split(/\s+/).length;
+  const wordCount2 = text.split(/\s+/).length;
   if (task.urgency === "critical")
     return "high";
-  if (wordCount > 150 || task.successCriteria.length > 5)
+  if (wordCount2 > 150 || task.successCriteria.length > 5)
     return "high";
-  if (wordCount > 60 || task.successCriteria.length > 2)
+  if (wordCount2 > 60 || task.successCriteria.length > 2)
     return "medium";
   return "low";
 }
@@ -265643,8 +265643,8 @@ function classifyQuery(query) {
     return "error";
   if (/^[A-Za-z_$][A-Za-z0-9_$]*$/.test(trimmed))
     return "symbol";
-  const wordCount = trimmed.split(/\s+/).length;
-  if (wordCount <= 3)
+  const wordCount2 = trimmed.split(/\s+/).length;
+  if (wordCount2 <= 3)
     return "short";
   return "long";
 }
@@ -270703,10 +270703,10 @@ Integrate this guidance into your current approach. Continue working on the task
           const turnTier = this.options.modelTier ?? "large";
           if (turn === 0 && (turnTier === "small" || turnTier === "medium")) {
             const goal = this._taskState.goal || "";
-            const wordCount = goal.split(/\s+/).length;
+            const wordCount2 = goal.split(/\s+/).length;
             const hasMultipleActions = /\band\b.*\band\b|then.*then|also.*also/i.test(goal);
             const hasMultipleFiles = /files?.*files?|\.ts.*\.ts|create.*write|modify.*create/i.test(goal);
-            const isComplex = wordCount > 40 || hasMultipleActions || hasMultipleFiles;
+            const isComplex = wordCount2 > 40 || hasMultipleActions || hasMultipleFiles;
             if (isComplex) {
               messages2.push({
                 role: "user",
@@ -294676,6 +294676,8 @@ var init_voice = __esm({
       currentPlayback = null;
       speakQueue = [];
       speaking = false;
+      drainPromise = null;
+      drainResolve = null;
       phonemizeFn = null;
       /** True when current model uses MLX Audio backend */
       mlxActive = false;
@@ -294992,6 +294994,15 @@ var init_voice = __esm({
         const speedFactor = emotion ? emotionToSpeedFactor(emotion, this.starkMode, this.autistMode) : 1;
         this.enqueueSpeech(text, 0.55, 1 + pitchBias, speedFactor, 0.15);
       }
+      /** Wait until the speak queue is fully drained (all audio played). */
+      async waitUntilIdle() {
+        if (!this.speaking && this.speakQueue.length === 0) return;
+        if (this.drainPromise) {
+          await this.drainPromise;
+        } else {
+          await this.sleep(100);
+        }
+      }
       enqueueSpeech(text, volume, pitchFactor, speedFactor = 1, stereoDelayMs = 0.6) {
         if (!this.enabled || !this.ready) return;
         text = sanitizeForTTS(text);
@@ -295190,6 +295201,11 @@ var init_voice = __esm({
        */
       async drainQueue() {
         this.speaking = true;
+        if (!this.drainPromise) {
+          this.drainPromise = new Promise((resolve40) => {
+            this.drainResolve = resolve40;
+          });
+        }
         let isFirst = true;
         let prefetchedWav = null;
         while (this.speakQueue.length > 0) {
@@ -295241,6 +295257,14 @@ var init_voice = __esm({
           }
         }
         this.speaking = false;
+        if (this.drainResolve) {
+          try {
+            this.drainResolve();
+          } catch {
+          }
+        }
+        this.drainResolve = null;
+        this.drainPromise = null;
       }
       sleep(ms) {
         return new Promise((resolve40) => setTimeout(resolve40, ms));
@@ -328218,15 +328242,62 @@ __export(voicechat_exports, {
   VoiceChatSession: () => VoiceChatSession
 });
 import { EventEmitter as EventEmitter10 } from "node:events";
-var VAD_SILENCE_MS, MAX_SEGMENT_MS, SUMMARY_INJECTION_INTERVAL, MAX_CONTEXT_TURNS, SYSTEM_PROMPT2, VoiceChatSession;
+function clamp01(x) {
+  return x < 0 ? 0 : x > 1 ? 1 : x;
+}
+function alnumRatio(s2) {
+  if (!s2) return 0;
+  const al = (s2.match(/[\p{L}\p{N}]/gu) || []).length;
+  return al / s2.length;
+}
+function wordCount(s2) {
+  const words = s2.trim().match(/[\p{L}\p{N}][\p{L}\p{N}'’_-]*/gu);
+  return words ? words.length : 0;
+}
+function repeatingCharPenalty(s2) {
+  let maxRun = 1, cur = 1;
+  for (let i2 = 1; i2 < s2.length; i2++) {
+    if (s2[i2] === s2[i2 - 1]) cur++;
+    else {
+      if (cur > maxRun) maxRun = cur;
+      cur = 1;
+    }
+  }
+  if (cur > maxRun) maxRun = cur;
+  return Math.min(1, Math.max(0, (maxRun - 3) / 10));
+}
+function computeSignalFromText(text, confidence) {
+  const t2 = text.trim();
+  if (!t2) return 0;
+  if (NOISE_ONLY_RE.test(t2)) return 0.05;
+  const len = t2.length;
+  const wc = wordCount(t2);
+  const alpha = alnumRatio(t2);
+  let score = 0;
+  if (wc >= 6 && alpha >= 0.6) score = 0.85;
+  else if (wc >= 3 && alpha >= 0.5) score = 0.7;
+  else if (wc >= 2 && alpha >= 0.4) score = 0.5;
+  else if (wc >= 1 && alpha >= 0.3 && len >= 4) score = 0.35;
+  else score = 0.15;
+  score -= repeatingCharPenalty(t2) * 0.4;
+  if (typeof confidence === "number" && !Number.isNaN(confidence)) {
+    score = 0.7 * score + 0.3 * clamp01(confidence);
+  }
+  return clamp01(score);
+}
+function truncateForLog(s2, n2) {
+  return s2.length <= n2 ? s2 : s2.slice(0, n2 - 1) + "…";
+}
+var VAD_SILENCE_MS, MAX_SEGMENT_MS, MAX_CONTEXT_TURNS, SYSTEM_PROMPT2, MIN_SIGNAL_SCORE, NOISE_ONLY_RE, VoiceChatSession;
 var init_voicechat = __esm({
   "packages/cli/src/tui/voicechat.ts"() {
     "use strict";
-    VAD_SILENCE_MS = 1100;
+    VAD_SILENCE_MS = 2e3;
     MAX_SEGMENT_MS = 6500;
-    SUMMARY_INJECTION_INTERVAL = 4;
     MAX_CONTEXT_TURNS = 20;
     SYSTEM_PROMPT2 = `You are a voice assistant having a live spoken conversation. Keep responses extremely brief — 1-2 sentences max. You're speaking aloud, not writing. Be conversational, direct, and helpful. Don't use markdown, bullet points, or formatting — just natural speech. If you don't know something, say so briefly. Do not over-think — respond quickly and concisely.`;
+    MIN_SIGNAL_SCORE = 0.4;
+    NOISE_ONLY_RE = /^(?:[.·…\s,;:!?\-–—_()\[\]{}"'`]+|(?:uh|um|erm|hmm|mm+|uhh+|umm+)[\s.!?]*)+$/i;
     VoiceChatSession = class extends EventEmitter10 {
       voice;
       listen;
@@ -328245,6 +328316,7 @@ var init_voicechat = __esm({
       captureStartTime = 0;
       silenceTimer = null;
       maxSegmentTimer = null;
+      lastSignalScore = null;
       // Abort control for inference
       abortController = null;
       // Callbacks
@@ -328256,6 +328328,7 @@ var init_voicechat = __esm({
       // Bound handlers for cleanup
       _onTranscript = null;
       _onError = null;
+      _retryMicTimer = null;
       constructor(opts) {
         super();
         this.voice = opts.voice;
@@ -328307,20 +328380,37 @@ var init_voicechat = __esm({
         this._onTranscript = (...args) => {
           let text;
           let isFinal;
+          let snr;
+          let confidence;
           if (typeof args[0] === "object" && args[0] !== null) {
             const evt = args[0];
             text = evt.text ?? "";
             isFinal = evt.isFinal ?? false;
+            snr = evt.snr;
+            confidence = evt.confidence;
           } else {
             text = String(args[0] ?? "");
             isFinal = Boolean(args[1]);
           }
           if (!text.trim()) return;
-          this.handleTranscript(text.trim(), isFinal);
+          this.handleTranscript(text.trim(), isFinal, snr, confidence);
         };
         this._onError = (err) => {
           const msg = err instanceof Error ? err.message : String(err);
           this.onStatus(`ASR error (voicechat continues without mic): ${msg.slice(0, 80)}`);
+          if (this.active && !this._retryMicTimer) {
+            this._retryMicTimer = setTimeout(async () => {
+              this._retryMicTimer = null;
+              if (!this.active) return;
+              try {
+                await this.listen.stop().catch(() => {
+                });
+                await this.listen.start();
+                this.onStatus("Mic auto-recovered — LISTENING");
+              } catch {
+              }
+            }, 1e3);
+          }
         };
         this.listen.on("transcript", this._onTranscript);
         this.listen.on("error", this._onError);
@@ -328372,7 +328462,7 @@ var init_voicechat = __esm({
       // ---------------------------------------------------------------------------
       // Transcript handling — VAD-style segment capture (Voryn pattern)
       // ---------------------------------------------------------------------------
-      handleTranscript(text, isFinal) {
+      handleTranscript(text, isFinal, snr, confidence) {
         if (!this.active) return;
         if (this._state !== "LISTENING" && this._state !== "CAPTURING") {
           return;
@@ -328388,6 +328478,8 @@ var init_voicechat = __esm({
           }, MAX_SEGMENT_MS);
         }
         this.captureBuffer = text;
+        this.lastSignalScore = typeof snr === "number" && !Number.isNaN(snr) ? clamp01(snr) : computeSignalFromText(text, confidence);
+        this.emit("snr", { score: this.lastSignalScore });
         this.onPartialTranscript(text);
         if (this.silenceTimer) clearTimeout(this.silenceTimer);
         if (isFinal) {
@@ -328418,10 +328510,25 @@ var init_voicechat = __esm({
           this.setState("LISTENING");
           return;
         }
+        const score = this.lastSignalScore ?? computeSignalFromText(text);
+        if (score < MIN_SIGNAL_SCORE || NOISE_ONLY_RE.test(text)) {
+          this.onStatus(`Ignoring low-signal utterance (SNR:${score.toFixed(2)}): ${truncateForLog(text, 48)}`);
+          this.emit("snrFiltered", { score, text });
+          this.setState("LISTENING");
+          this.captureBuffer = "";
+          this.lastSignalScore = null;
+          return;
+        }
         this.setState("TRANSCRIBING");
         this.onUserSpeech(text);
         this.context.push({ role: "user", content: text });
         this.turnCount++;
+        if (this.runner) {
+          try {
+            this.runner.injectUserMessage(`[VOICECHAT] ${text}`);
+          } catch {
+          }
+        }
         while (this.context.length > MAX_CONTEXT_TURNS + 1) {
           this.context.splice(1, 1);
         }
@@ -328443,11 +328550,18 @@ var init_voicechat = __esm({
             this.setState("SPEAKING");
             this.onAgentSpeech(response.trim());
             this.voice.speak(response.trim());
-            if (this.runner && this.turnCount % SUMMARY_INJECTION_INTERVAL === 0) {
+            if (this.runner) {
               this.injectSummary();
             }
-            const estimatedMs = Math.max(1500, response.length / 5 * (6e4 / 150));
-            await new Promise((r2) => setTimeout(r2, estimatedMs));
+            if (typeof this.voice.waitUntilIdle === "function") {
+              try {
+                await this.voice.waitUntilIdle();
+              } catch {
+              }
+            } else {
+              const estimatedMs = Math.max(1500, response.length / 5 * (6e4 / 150));
+              await new Promise((r2) => setTimeout(r2, estimatedMs));
+            }
           }
         } catch (err) {
           if (!this.active) return;
@@ -328570,13 +328684,19 @@ var init_voicechat = __esm({
       // ---------------------------------------------------------------------------
       injectSummary() {
         if (!this.runner) return;
-        const recentTurns = this.context.filter((t2) => t2.role !== "system").slice(-6).map((t2) => `${t2.role === "user" ? "User" : "Assistant"}: ${t2.content}`).join("\n");
+        const recentTurns = this.context.filter((t2) => t2.role !== "system").slice(-8).map((t2) => `${t2.role === "user" ? "User" : "Assistant"}: ${t2.content}`).join("\n");
         this.runner.injectUserMessage(
-          `[VOICECHAT SUMMARY] The following is a summary of the recent voice conversation happening in parallel. You don't need to respond to this directly — it's for your awareness. Continue your current task.
+          `[VOICECHAT SUMMARY] Parallel voice liaison update (for awareness only). Continue your current task; do not respond to this directly.
 
 ${recentTurns}`
         );
       }
+      /** Enqueue narration from main agent events into the voice channel */
+      enqueueAgentNarration(text, subordinate = true) {
+        if (!text || !this.active) return;
+        if (subordinate) this.voice.speakSubordinate(text);
+        else this.voice.speak(text);
+      }
     };
   }
 });
@@ -330156,6 +330276,17 @@ ${entry.fullContent}`
             }
           });
         }
+        if (voice?.enabled && voice.voiceMode === "voicechat" && _voiceChatSession?.isActive && event.toolName === "task_complete") {
+          const emoStateFinal = emotionEngine?.getState();
+          const emoCtxFinal = emoStateFinal ? { valence: emoStateFinal.valence, arousal: emoStateFinal.arousal, label: emoStateFinal.label, emoji: emoStateFinal.emoji } : void 0;
+          const desc = describeTaskComplete(String(event.content ?? ""), true, vLevel);
+          if (desc) {
+            try {
+              _voiceChatSession.enqueueAgentNarration(desc, false);
+            } catch {
+            }
+          }
+        }
         break;
       }
       case "model_response":
@@ -330237,6 +330368,15 @@ ${entry.fullContent}`
             voice.speak(chatText);
             voice._spokenStreamText = true;
           }
+        } else if (voice?.enabled && voice.voiceMode === "voicechat" && (streamTextBuffer || event.content)) {
+          const chatText = (streamTextBuffer || event.content || "").trim();
+          streamTextBuffer = "";
+          if (chatText.length > 10 && _voiceChatSession?.isActive) {
+            try {
+              _voiceChatSession.enqueueAgentNarration(chatText, false);
+            } catch {
+            }
+          }
         }
         break;
       }
@@ -331274,7 +331414,7 @@ ${opts.systemPromptAddition}` : `Working directory: ${repoRoot}`;
   autoUpdateTimer.unref();
   const voiceEngine = new VoiceEngine();
   let voiceSession = null;
-  let _voiceChatSession = null;
+  let _voiceChatSession2 = null;
   let tunnelGateway = null;
   let p2pGateway = null;
   let peerMesh = null;
@@ -332963,7 +333103,7 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
     },
     // --- /voicechat: Voryn-style state machine voice conversation ---
     async voiceChatStart() {
-      if (_voiceChatSession?.isActive) return;
+      if (_voiceChatSession2?.isActive) return;
       if (!voiceEngine.enabled || !voiceEngine.ready) {
         writeContent(() => renderInfo2("Auto-enabling voice for voice chat..."));
         const voiceMsg = await voiceEngine.toggle();
@@ -332980,7 +333120,7 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
           }
         }
       };
-      _voiceChatSession = new VoiceChatSession2({
+      _voiceChatSession2 = new VoiceChatSession2({
         voice: voiceEngine,
         listen: listenEng,
         backendUrl: currentConfig.backendUrl,
@@ -333005,16 +333145,28 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
           writeContent(() => renderInfo2(`\x1B[38;5;243m[voicechat] ${state}\x1B[0m`));
         }
       });
-      await _voiceChatSession.start();
+      _voiceChatSession2.on("snr", (e2) => {
+        const s2 = typeof e2?.score === "number" ? Math.max(0, Math.min(1, e2.score)) : null;
+        if (s2 !== null) {
+          writeContent(() => {
+            process.stdout.write(`\r\x1B[2K\x1B[38;5;243m  [hearing] (snr:${s2.toFixed(2)})\x1B[0m`);
+          });
+        }
+      });
+      _voiceChatSession2.on("snrFiltered", (e2) => {
+        const s2 = typeof e2?.score === "number" ? e2.score.toFixed(2) : "?";
+        writeContent(() => renderInfo2(`\x1B[38;5;243m[voicechat]\x1B[0m dropped low-signal utterance (SNR:${s2})`));
+      });
+      await _voiceChatSession2.start();
     },
     async voiceChatStop() {
-      if (_voiceChatSession?.isActive) {
-        await _voiceChatSession.stop();
-        _voiceChatSession = null;
+      if (_voiceChatSession2?.isActive) {
+        await _voiceChatSession2.stop();
+        _voiceChatSession2 = null;
       }
     },
     isVoiceChatActive() {
-      return _voiceChatSession?.isActive ?? false;
+      return _voiceChatSession2?.isActive ?? false;
     },
     async exposeStart(kindOrUrl, authKey, transport, fullAccess, passthrough, loadbalance) {
       const knownKinds = ["ollama", "vllm", "llvm", "passthrough"];

package/dist/launcher.cjs

@@ -1,14 +1,78 @@
 #!/usr/bin/env node
-// Minimal launcher shim for open-agents CLI — forwards to bundled index.js
-// Use dynamic import() so ESM with top-level await can load correctly.
-(async () => {
+// Robust launcher for open-agents CLI.
+// - Runs the ESM entry as a child process
+// - On exit code 120 (update), resets terminal and restarts child
+// - Prevents raw-mode/mouse-tracking bleedthrough on restart or crash
+
+const { spawn, spawnSync } = require('node:child_process');
+const { resolve } = require('node:path');
+
+function resetTerminal() {
+  try { if (process.stdin.isTTY && typeof process.stdin.setRawMode === 'function') process.stdin.setRawMode(false); } catch {}
+  // Disable mouse tracking, bracketed paste, show cursor, reset attrs, exit alt screen if active
+  const ESC = '\x1B';
   try {
-    const { resolve } = require('node:path');
-    const { pathToFileURL } = require('node:url');
-    const entryUrl = pathToFileURL(resolve(__dirname, 'index.js')).href;
-    await import(entryUrl);
-  } catch (e) {
-    console.error('Failed to launch open-agents:', e && e.message ? e.message : e);
-    process.exit(1);
+    process.stdout.write(
+      ESC + '[?25h' +             // show cursor
+      ESC + '[?1000l' +           // X10 mouse off
+      ESC + '[?1002l' +           // button-event mouse off
+      ESC + '[?1003l' +           // any-event mouse off
+      ESC + '[?1006l' +           // SGR mouse off
+      ESC + '[?1015l' +           // urxvt mouse off
+      ESC + '[?2004l' +           // bracketed paste off
+      ESC + '[?1049l' +           // exit alt screen
+      ESC + '[0m'                 // reset attributes
+    );
+  } catch {}
+  // stty sane (POSIX)
+  if (process.platform !== 'win32' && process.stdin.isTTY) {
+    try { spawnSync('stty', ['sane'], { stdio: 'inherit' }); } catch {}
+  }
+}
+
+function runChild() {
+  const entry = resolve(__dirname, 'index.js');
+  const args = [entry, ...process.argv.slice(2)];
+  const child = spawn(process.execPath, args, {
+    stdio: 'inherit',
+    env: process.env,
+  });
+  return child;
+}
+
+(async () => {
+  let restarts = 0;
+  const MAX_RESTARTS = 3;
+  let child = runChild();
+
+  const forward = (sig) => {
+    try { child && child.kill(sig); } catch {}
+  };
+  process.on('SIGINT', () => forward('SIGINT'));
+  process.on('SIGTERM', () => forward('SIGTERM'));
+
+  function attach(childProc) {
+    childProc.on('exit', (code, signal) => {
+      if (signal) {
+        resetTerminal();
+        process.kill(process.pid, signal);
+        return;
+      }
+      if (code === 120 && restarts < MAX_RESTARTS) {
+        // Update-triggered restart
+        resetTerminal();
+        restarts += 1;
+        setTimeout(() => {
+          child = runChild();
+          attach(child);
+        }, 300);
+        return;
+      }
+      // Normal exit or too many restarts
+      resetTerminal();
+      process.exit(typeof code === 'number' ? code : 0);
+    });
   }
+
+  attach(child);
 })();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.277",
+  "version": "0.187.280",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",