agent.libx.js 0.93.1 → 0.93.4

package/cli/cli.ts

@@ -208,7 +208,7 @@ Flags:
                        impulsive reactions, human pacing (implies --duplex; aliases: --convo, --voice)
                        with SONIOX_API_KEY + CARTESIA_API_KEY(+VOICE_ID) set: real voice I/O — mic in,
                        spoken replies out (echo-cancelled; speak over it to interrupt)
-  --voice-model <id>   with --duplex: the fast voice model (default anthropic/claude-haiku-4-5)
+  --voice-model <id>   with --duplex: the fast voice model (default groq/openai/gpt-oss-20b)
   --add-dir <path>     mount another directory into the workspace (repeatable; disk mode only)
   --subagents          allow the Task tool (spawn child agents)
   --reasoning <e>      extended thinking: off|low|medium|high or a token budget (anthropic/openai)

package/dist/cli.js

@@ -3501,7 +3501,7 @@ var DuplexAgentOptions = class {
   ai;
   /** The WORKER's filesystem. If omitted the worker keeps Agent's jailed-disk-at-cwd default. */
   fs;
-  voiceModel = "anthropic/claude-haiku-4-5";
+  voiceModel = "groq/openai/gpt-oss-20b";
   workerModel = "anthropic/claude-sonnet-4-6";
   /** Escape hatches merged over the derived per-agent options. */
   voiceOptions;
@@ -4003,6 +4003,8 @@ var VoiceEngine = class {
   lastOverlapPartial = "";
   // change-detection: only NEW partial text counts as activity
   resumeTimer = null;
+  turnStartAt = 0;
+  // timestamp when the current turn began (for TTFT logging)
   constructor(options) {
     this.options = { ...new VoiceEngineOptions(), ...options };
     const o = this.options;
@@ -4063,6 +4065,7 @@ var VoiceEngine = class {
       this.spokeDeltas = true;
       this.ackAt = now();
     }
+    this.turnStartAt = now();
     this.setState("thinking");
   }
   speakDelta(text) {
@@ -4071,6 +4074,7 @@ var VoiceEngine = class {
     this.reply += text;
     for (const w of this.words(this.reply)) this.echoWords.add(w);
     this.tts.speak(text, true);
+    if (!this.spokeDeltas && this.turnStartAt) log7.info(`ttft: ${Math.round(now() - this.turnStartAt)}ms`);
     this.spokeDeltas = true;
     this.setState("speaking");
   }
@@ -4091,6 +4095,7 @@ var VoiceEngine = class {
       }
       this.drainTimer = null;
       this.speaking = false;
+      if (this.turnStartAt) log7.info(`turn: ${Math.round(now() - this.turnStartAt)}ms (incl. playback)`);
       this.echoUntil = now() + 2500;
       if (!this.usingAec) this.stt.reset();
       this.setState("listening");
@@ -4127,7 +4132,7 @@ var VoiceEngine = class {
     this.ctxOpen = false;
     this.interrupted = true;
     this.suspectUntil = 0;
-    this.echoUntil = now() + 2500;
+    this.echoUntil = now() + Math.max(2500, this.player.drainMs() + 3e3);
     this.tts.cancel();
     this.player.kill();
     if (!this.usingAec) this.stt.reset();
@@ -4322,6 +4327,8 @@ var SonioxSTT = class {
   lastChangeAt = 0;
   lastCombined = "";
   endpointTimer = null;
+  firstTokenAt = 0;
+  // first speech token in current utterance
   constructor(options) {
     this.options = { ...new SonioxSTTOptions(), ...options };
   }
@@ -4361,6 +4368,7 @@ var SonioxSTT = class {
     this.endpointTimer = setInterval(() => {
       const combined = (this.finalText + this.partialText).trim();
       if (!combined || now2() - this.lastChangeAt < this.options.silenceEndpointMs) return;
+      if (this.firstTokenAt) log8.info(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192silence-endpoint, "${combined.slice(0, 60)}"`);
       this.reset();
       this.onUtterance(combined, now2());
     }, 120);
@@ -4388,10 +4396,12 @@ var SonioxSTT = class {
     if (combined !== this.lastCombined) {
       this.lastCombined = combined;
       this.lastChangeAt = now2();
+      if (!this.firstTokenAt && combined.trim()) this.firstTokenAt = now2();
     }
     this.onPartial(combined);
     if (endpoint && this.finalText.trim()) {
       const utterance = this.finalText.trim();
+      if (this.firstTokenAt) log8.info(`stt: ${Math.round(now2() - this.firstTokenAt)}ms first-token\u2192endpoint, "${utterance.slice(0, 60)}"`);
       this.reset();
       this.onUtterance(utterance, now2());
     }
@@ -4400,6 +4410,7 @@ var SonioxSTT = class {
     this.finalText = "";
     this.partialText = "";
     this.lastCombined = "";
+    this.firstTokenAt = 0;
   }
   stop() {
     this.stopped = true;
@@ -4434,7 +4445,15 @@ var CartesiaTTS = class {
   constructor(options) {
     this.options = { ...new CartesiaTTSOptions(), ...options };
   }
+  closed = false;
+  connecting = null;
   async connect() {
+    this.closed = false;
+    this.connecting = this.doConnect();
+    await this.connecting;
+    this.connecting = null;
+  }
+  async doConnect() {
     const key = await resolveAuth(this.options.auth);
     const param = this.options.authMode === "token" ? "access_token" : "api_key";
     this.ws = new WebSocket(`wss://api.cartesia.ai/tts/websocket?cartesia_version=2026-03-01&${param}=${key}`);
@@ -4442,7 +4461,12 @@ var CartesiaTTS = class {
       this.ws.onopen = () => res();
       this.ws.onerror = (e) => rej(new Error(`cartesia ws: ${e.message || "connect failed"}`));
     });
-    this.ws.onclose = (ev) => log9.warn(`cartesia ws closed (${ev.code} ${ev.reason || ""})`);
+    this.ws.onclose = (ev) => {
+      log9.warn(`cartesia ws closed (${ev.code} ${ev.reason || ""})`);
+      if (!this.closed) {
+        this.connecting = this.doConnect().catch((e) => log9.error(`cartesia reconnect failed: ${e.message}`));
+      }
+    };
     this.ws.onmessage = (ev) => {
       const m = JSON.parse(String(ev.data));
       if (m.context_id && m.context_id !== this.ctxId) return;
@@ -4453,6 +4477,11 @@ var CartesiaTTS = class {
       else if (m.type === "error" && !/already been cancelled|does not exist/.test(m.message || "")) log9.warn(`cartesia: ${JSON.stringify(m)}`);
     };
   }
+  /** Ensure the WS is open before sending — reconnects if idle-closed. */
+  async ensureConnected() {
+    if (this.connecting) await this.connecting;
+    if (this.ws?.readyState !== WebSocket.OPEN) await this.connect();
+  }
   newContext() {
     this.ctxId = `ctx-${++this.ctxSeq}`;
     this.firstAudioAt = 0;
@@ -4470,6 +4499,7 @@ var CartesiaTTS = class {
   }
   speak(text, cont) {
     if (this.ws?.readyState === WebSocket.OPEN) this.ws.send(this.frame(text, cont));
+    else void this.ensureConnected().then(() => this.ws?.readyState === WebSocket.OPEN && this.ws.send(this.frame(text, cont)));
   }
   end() {
     if (this.ws?.readyState === WebSocket.OPEN) this.ws.send(this.frame("", false));
@@ -4478,6 +4508,7 @@ var CartesiaTTS = class {
     if (this.ws?.readyState === WebSocket.OPEN) this.ws.send(JSON.stringify({ context_id: this.ctxId, cancel: true }));
   }
   close() {
+    this.closed = true;
     if (this.ws) this.ws.onclose = null;
     this.ws?.close();
   }
@@ -7385,7 +7416,7 @@ Flags:
                        impulsive reactions, human pacing (implies --duplex; aliases: --convo, --voice)
                        with SONIOX_API_KEY + CARTESIA_API_KEY(+VOICE_ID) set: real voice I/O \u2014 mic in,
                        spoken replies out (echo-cancelled; speak over it to interrupt)
-  --voice-model <id>   with --duplex: the fast voice model (default anthropic/claude-haiku-4-5)
+  --voice-model <id>   with --duplex: the fast voice model (default groq/openai/gpt-oss-20b)
   --add-dir <path>     mount another directory into the workspace (repeatable; disk mode only)
   --subagents          allow the Task tool (spawn child agents)
   --reasoning <e>      extended thinking: off|low|medium|high or a token budget (anthropic/openai)