@stage-labs/metro 0.1.0-beta.237 → 0.1.0-beta.238

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@stage-labs/metro",
-  "version": "0.1.0-beta.237",
+  "version": "0.1.0-beta.238",
   "description": "The metro command line: run your agent on this machine, relaying its chat channels and connectors to Claude Code.",
   "license": "MIT",
   "repository": {

package/runtime/marketplace/plugin/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "metro",
-  "version": "0.1.0-beta.237",
+  "version": "0.1.0-beta.238",
   "description": "Metro for Claude Code, installed by the metro daemon on this machine: every connector the agent holds as its own MCP server relayed through the daemon, the orchestrator-only guard on the main thread, and the standing rules loaded at each session start. No vendor credential ever sits in a config file.",
   "author": {
     "name": "Bonustrack",

package/runtime/node_modules/@metro-labs/daemon/src/voice/call.ts

@@ -111,6 +111,7 @@ export class Call {
       },
     });
     const sdp = await this.peer.answer(offer);
+    if (this.over) return;
     await this.signal({ kind: 'answer', callId: this.start.callId, from: this.selfPeer, to, sdp });
   }
 
@@ -147,8 +148,9 @@ export class Call {
     this.peer?.close();
     const notes = await this.talk.finish();
     const minutes = Math.max(1, Math.round((Date.now() - this.startedAt) / 60_000));
-    if (notes !== null) await trainCall('send', { line: this.start.line, text: `Call notes (${String(minutes)} min):\n${notes}` });
-    else if (!reason.startsWith('the caller') && reason !== 'no offer arrived')
-      await trainCall('send', { line: this.start.line, text: `The voice call ended: ${reason}.` });
+    const timing = this.talk.replyTime === null ? '' : `\n${this.talk.replyTime}`;
+    const failure = reason.startsWith('the caller') || reason === 'no offer arrived' ? null : `The voice call ended: ${reason}.`;
+    const text = [failure, notes === null ? null : `Call notes (${String(minutes)} min):\n${notes}${timing}`].filter((part) => part !== null).join('\n\n');
+    if (text !== '') await trainCall('send', { line: this.start.line, text });
   }
 }

package/runtime/node_modules/@metro-labs/daemon/src/voice/speech.ts

@@ -26,14 +26,19 @@ const charsOf = (msg: Record<string, unknown>): number => {
   return isRecord(alignment) && Array.isArray(alignment.chars) ? alignment.chars.length : 0;
 };
 
+const MAX_REASON = 300;
+
 export class Utterance {
   private readonly ws: WebSocket;
   private readonly outbox: string[] = [];
   private listener: Voiced | null = null;
   private carry = 0;
   private closed = false;
+  private aborted = false;
+  private voiced = false;
+  private refusal: string | null = null;
 
-  constructor(apiKey: string, voiceId: string) {
+  constructor(apiKey: string, voiceId: string, failed: (reason: string) => void) {
     this.ws = new WebSocket(ttsUrl(voiceId), { headers: { 'xi-api-key': apiKey } });
     this.ws.on('open', () => {
       this.ws.send(JSON.stringify({ text: ' ' }));
@@ -43,12 +48,15 @@ export class Utterance {
       this.onMessage(wsText(data));
     });
     this.ws.on('error', (err) => {
+      this.refusal ??= errMsg(err);
       log.warn({ err: errMsg(err) }, 'voice: text to speech connection error');
     });
-    this.ws.on('close', () => {
+    this.ws.on('close', (code, why) => {
       this.closed = true;
       this.listener?.done();
       this.listener = null;
+      const reason = this.refusal ?? (code === 1000 ? null : (why.toString() || `closed (${String(code)})`));
+      if (!this.aborted && !this.voiced && reason !== null) failed(reason.slice(0, MAX_REASON));
     });
   }
 
@@ -69,6 +77,7 @@ export class Utterance {
   }
 
   abort(): void {
+    this.aborted = true;
     this.listener = null;
     this.ws.terminate();
   }
@@ -82,10 +91,14 @@ export class Utterance {
   private onMessage(raw: string): void {
     const msg = parsed(raw);
     if (typeof msg.audio === 'string' && msg.audio !== '') {
+      this.voiced = true;
       const pcm = pcmOf(Buffer.from(msg.audio, 'base64'));
       const up = upsample2(pcm, this.carry);
       this.carry = pcm.at(-1) ?? this.carry;
       this.listener?.audio(up, charsOf(msg));
-    } else if (msg.isFinal !== true) log.warn({ msg: raw.slice(0, 300) }, 'voice: text to speech said');
+    } else if (msg.isFinal !== true) {
+      if (typeof msg.message === 'string' && msg.message !== '') this.refusal ??= msg.message;
+      log.warn({ msg: raw.slice(0, MAX_REASON) }, 'voice: text to speech said');
+    }
   }
 }

package/runtime/node_modules/@metro-labs/daemon/src/voice/talk.ts

@@ -6,7 +6,7 @@ import { Utterance } from './speech.js';
 import { Chunker } from './speakable.js';
 import { brainModel, voiceIdOf, type VoiceConfig } from './store.js';
 
-const BARGE_MIN_CHARS = 4;
+const BARGE_MIN_WORDS = 2;
 const NOTES_WAIT_MS = 60_000;
 const NOTES_PROMPT =
   '[The call has ended. Write short call notes for the chat: what was discussed, what was decided and what you will follow up on. Plain text, at most 6 short lines, no title and no greeting. Reply with the notes only.]';
@@ -17,10 +17,14 @@ interface Reply {
   heard: number;
   chunker: Chunker;
   askedAt: number | null;
+  saidAt: number | null;
   firstText: number | null;
   firstAudio: number | null;
+  firstPlayed: number | null;
 }
 
+const wordsIn = (text: string): number => text.split(/\s+/).filter((word) => /\p{L}/u.test(word)).length;
+
 const since = (start: number | null, at: number | null): number | null =>
   start === null || at === null ? null : Math.round(at - start);
 
@@ -43,11 +47,15 @@ export class Talk implements Thinking {
   private muted = false;
   private busy = false;
   private askedAt: number | null = null;
+  private saidAt: number | null = null;
+  private lastHeardAt: number | null = null;
+  private readonly delays: number[] = [];
   private notes: ((text: string) => void) | null = null;
   private notesStarted = false;
   private live = false;
   private finished = false;
   private heardCaller = false;
+  private spoke = false;
 
   constructor(
     private readonly cfg: VoiceConfig,
@@ -76,7 +84,7 @@ export class Talk implements Thinking {
         this.ended(reason);
       },
     });
-    this.spare = new Utterance(cfg.apiKey, voiceIdOf(cfg));
+    this.spare = this.utterance();
   }
 
   prime(context: string): void {
@@ -86,7 +94,7 @@ export class Talk implements Thinking {
 
   connect(): void {
     this.live = true;
-    this.spare ??= new Utterance(this.cfg.apiKey, voiceIdOf(this.cfg));
+    this.spare ??= this.utterance();
     this.player.start();
   }
 
@@ -115,8 +123,19 @@ export class Talk implements Thinking {
     this.busy = true;
     if (this.notes !== null) this.notesStarted = true;
     if (this.notes !== null || this.finished) return;
-    this.reply = { utterance: this.takeUtterance(), said: '', heard: 0, chunker: new Chunker(), askedAt: this.askedAt, firstText: null, firstAudio: null };
+    this.reply = {
+      utterance: this.takeUtterance(),
+      said: '',
+      heard: 0,
+      chunker: new Chunker(),
+      askedAt: this.askedAt,
+      saidAt: this.saidAt,
+      firstText: null,
+      firstAudio: null,
+      firstPlayed: null,
+    };
     this.askedAt = null;
+    this.saidAt = null;
   }
 
   text(delta: string): void {
@@ -158,14 +177,15 @@ export class Talk implements Thinking {
   }
 
   private takeUtterance(): Utterance {
-    const ready = this.spare?.usable === true ? this.spare : new Utterance(this.cfg.apiKey, voiceIdOf(this.cfg));
-    this.spare = this.live ? new Utterance(this.cfg.apiKey, voiceIdOf(this.cfg)) : null;
+    const ready = this.spare?.usable === true ? this.spare : this.utterance();
+    this.spare = this.live ? this.utterance() : null;
     if (ready === this.spare) this.spare = null;
     const reply = (): Reply | null => this.reply;
     ready.attach({
       audio: (pcm, chars) => {
         const current = reply();
         if (current?.utterance !== ready) return;
+        this.spoke = true;
         current.firstAudio ??= performance.now();
         this.player.push(pcm, chars);
       },
@@ -176,6 +196,13 @@ export class Talk implements Thinking {
     return ready;
   }
 
+  private utterance(): Utterance {
+    return new Utterance(this.cfg.apiKey, voiceIdOf(this.cfg), (reason) => {
+      log.warn({ reason }, 'voice: text to speech gave no audio');
+      if (!this.spoke && !this.finished) this.ended(`ElevenLabs could not speak (${reason})`);
+    });
+  }
+
   private speak(reply: Reply, parts: string[]): void {
     for (const part of parts) {
       reply.said += `${part} `;
@@ -187,32 +214,46 @@ export class Talk implements Thinking {
     if (this.reply !== null) this.reply.heard += chars;
   }
 
+  get replyTime(): string | null {
+    if (this.delays.length === 0) return null;
+    const sorted = [...this.delays].sort((a, b) => a - b);
+    const middle = sorted[Math.floor(sorted.length / 2)] ?? 0;
+    return `Reply time, from your last word heard to my first word sent: about ${(middle / 1000).toFixed(1)} s (middle of ${String(sorted.length)}).`;
+  }
+
   private firstPlayed(): void {
     const reply = this.reply;
-    if (reply?.askedAt != null) log.info({ firstPlayed: since(reply.askedAt, performance.now()) }, 'voice: first reply audio sent to the caller');
+    if (reply?.firstPlayed !== null || reply.saidAt === null) return;
+    reply.firstPlayed = performance.now();
+    this.delays.push(reply.firstPlayed - reply.saidAt);
+    log.info({ fromLastWord: since(reply.saidAt, reply.firstPlayed), fromCommit: since(reply.askedAt, reply.firstPlayed) }, 'voice: first reply audio sent to the caller, in ms');
   }
 
-  private partial(text: string): void {
-    if (text.replace(/\s/g, '').length < BARGE_MIN_CHARS || this.muted) return;
-    if (!this.player.busy && !this.busy) return;
-    const reply = this.reply;
-    this.cut = heardNote(reply);
-    reply?.utterance.abort();
+  private interrupts(text: string): boolean {
+    return !this.muted && this.player.busy && wordsIn(text) >= BARGE_MIN_WORDS;
+  }
+
+  private cutOff(): void {
+    this.cut = heardNote(this.reply);
+    this.reply?.utterance.abort();
     this.player.clear();
     this.muted = true;
   }
 
+  private partial(text: string): void {
+    if (text !== '') this.lastHeardAt = performance.now();
+    if (this.interrupts(text)) this.cutOff();
+  }
+
   private committed(text: string): void {
     this.heardCaller = true;
+    if (this.interrupts(text)) this.cutOff();
     const note = this.cut;
-    const now = this.busy || note !== null;
+    const now = this.muted;
     this.cut = null;
-    if (now && !this.muted) {
-      this.reply?.utterance.abort();
-      this.player.clear();
-      this.muted = true;
-    }
     this.askedAt = performance.now();
+    this.saidAt = this.lastHeardAt ?? this.askedAt;
+    this.lastHeardAt = null;
     this.busy = true;
     this.brain.tell(note === null ? text : `${note}\n${text}`, now);
   }

package/runtime/runtime.json

@@ -1,3 +1,3 @@
 {
-  "version": "0.1.0-beta.237"
+  "version": "0.1.0-beta.238"
 }