agent.libx.js 0.93.2 → 0.93.6

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

@@ -2900,9 +2900,17 @@ var Agent = class _Agent {
       let res;
       const sent = this.trimContext();
       const frag = reasoningToChatFragment(o.model, o.reasoning);
+      const isCursorWithTools = o.model.startsWith("cursor/") && wireTools.length > 0;
+      const cursorPo = isCursorWithTools ? {
+        toolExecutor: async (name, args) => {
+          const tc = { id: `cursor-${Date.now()}`, type: "function", function: { name, arguments: JSON.stringify(args) } };
+          const raw = await this.dispatch(tc);
+          return typeof raw === "string" ? raw : raw.text;
+        }
+      } : void 0;
       const reasonOpts = {
         ...frag,
-        ...o.providerOptions ? { providerOptions: { ...frag.providerOptions, ...o.providerOptions } } : {}
+        ...o.providerOptions || cursorPo ? { providerOptions: { ...frag.providerOptions, ...o.providerOptions, ...cursorPo } } : {}
       };
       try {
         if (useStream) {
@@ -3501,7 +3509,7 @@ 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 +4011,8 @@ 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 +4073,7 @@ var VoiceEngine = class {
       this.spokeDeltas = true;
       this.ackAt = now();
     }
+    this.turnStartAt = now();
     this.setState("thinking");
   }
   speakDelta(text) {
@@ -4071,6 +4082,7 @@ 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 +4103,7 @@ 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 +4140,7 @@ 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 +4335,8 @@ 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 +4376,7 @@ 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 +4404,12 @@ 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 +4418,7 @@ var SonioxSTT = class {
     this.finalText = "";
     this.partialText = "";
     this.lastCombined = "";
+    this.firstTokenAt = 0;
   }
   stop() {
     this.stopped = true;
@@ -7405,7 +7424,7 @@ 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)