@ouro.bot/cli 0.1.0-alpha.592 → 0.1.0-alpha.595

package/changelog.json

@@ -1,13 +1,28 @@
 {
   "_note": "This changelog is maintained as part of the PR/version-bump workflow. Agent-curated, not auto-generated. Agents read this file directly via read_file to understand what changed between versions.",
   "versions": [
+    {
+      "version": "0.1.0-alpha.595",
+      "changes": [
+        "Voice phone transport defaults to media-stream when OpenAI Realtime or OpenAI SIP is configured but no explicit voice.twilioTransportMode is set. Previously the default was record-play, which made conversationEngine resolve to cascade and routed inbound calls through the ElevenLabs/Whisper greeting path operators with realtime-only credentials never configured — producing a fully silent first turn (\"no greeting at all\"). Realtime requires media-stream by nature, so we now infer it. Defensive prewarm guard branch marked with a v8 ignore since the implicit default makes it unreachable in current outbound tests."
+      ]
+    },
+    {
+      "version": "0.1.0-alpha.594",
+      "changes": [
+        "Voice phone transport defaults to media-stream when OpenAI Realtime or OpenAI SIP is configured but no explicit voice.twilioTransportMode is set. Previously the default was record-play, which made conversationEngine resolve to cascade and routed inbound calls through the ElevenLabs/Whisper greeting path operators with realtime-only credentials never configured — producing a fully silent first turn (\"no greeting at all\"). Realtime requires media-stream by nature, so we now infer it."
+      ]
+    },
+    {
+      "version": "0.1.0-alpha.593",
+      "changes": [
+        "Voice runtime now emits assistant.speech.cancelled to the floor when the caller barges in (previously the floor was left with floorOwner=caller permanently after the interrupted greeting's response.done arrived, so Slugger went silent for the rest of every barged-in call). Also wires input_audio_buffer.speech_stopped to apply caller.speech.ended so VAD-detected sub-vocal sounds that never transcribe cannot strand the floor. No live human calls."
+      ]
+    },
     {
       "version": "0.1.0-alpha.592",
       "changes": [
-        "Local mail store (`FileMailroomStore`) now writes messages and raw RFC822 bodies in plaintext on disk. The agent bundle becomes the durable archive surface for mail — a vault key rotation or lost private key can no longer render the local archive unreadable, and `ouro mail import-mbox` remains the canonical recovery mechanism.",
-        "Cloud-blob mail store (`AzureBlobMailroomStore`) continues to encrypt every message on the wire and at rest. The trust boundary it crosses is real; nothing about the hosted path changes.",
-        "`StoredMailMessage` is now a discriminated union on `bodyForm`: plaintext messages carry `private` inline, encrypted messages carry `privateEnvelope`. All reader code paths converge on a single `readPrivateEnvelope` accessor that branches on `bodyForm` so callers no longer reach into decrypt internals.",
-        "One-shot migration helper runs on `FileMailroomStore` init: wipes pre-change encrypted-shape residue in `messages/` and `raw/`, deletes stale `azure-blob` coverage records, and prunes orphan search-cache documents that no longer reference a present local message. Idempotent — subsequent runs are no-ops."
+        "Voice runtime now emits assistant.speech.cancelled to the floor when the caller barges in (previously the floor was left with floorOwner=caller permanently after the interrupted greeting's response.done arrived, so Slugger went silent for the rest of every barged-in call). Also wires input_audio_buffer.speech_stopped to apply caller.speech.ended, so VAD-detected sub-vocal sounds that never transcribe cannot strand the floor. No live human calls."
       ]
     },
     {

package/dist/senses/voice/twilio-phone-runtime.js

@@ -248,8 +248,20 @@ function resolveTwilioPhoneTransportRuntime(options) {
         ?? configString(options.machineConfig, "voice.twilioBasePath")
         ?? options.defaultBasePath
         ?? twilio_phone_1.TWILIO_PHONE_WEBHOOK_BASE_PATH);
+    const explicitTransportModeString = configString(options.machineConfig, "voice.twilioTransportMode");
+    // When the operator has only configured OpenAI Realtime (key) or OpenAI SIP
+    // (project id) and not picked a transport mode, infer media-stream — the
+    // legacy `record-play` default would otherwise pin `conversationEngine` to
+    // `cascade`, route inbound calls through the ElevenLabs/Whisper greeting
+    // path the operator never configured, and produce a fully silent first
+    // turn ("no greeting at all"). Realtime requires media-stream by nature.
+    const hasRealtimeApiKey = !!resolveOpenAIRealtimeApiKey({ runtimeConfig: options.runtimeConfig, overrides });
+    const hasSipProjectConfig = !!(configString(options.runtimeConfig, "voice.openaiSipProjectId")
+        || configString(options.machineConfig, "voice.openaiSipProjectId"));
+    const realtimeImpliesMediaStream = hasRealtimeApiKey || hasSipProjectConfig;
     const transportMode = overrides.transportMode
-        ?? (0, twilio_phone_1.normalizeTwilioPhoneTransportMode)(configString(options.machineConfig, "voice.twilioTransportMode") ?? twilio_phone_1.DEFAULT_TWILIO_PHONE_TRANSPORT_MODE);
+        ?? (0, twilio_phone_1.normalizeTwilioPhoneTransportMode)(explicitTransportModeString
+            ?? (realtimeImpliesMediaStream ? "media-stream" : twilio_phone_1.DEFAULT_TWILIO_PHONE_TRANSPORT_MODE));
     const conversationEngine = configuredConversationEngine(options, overrides, transportMode);
     const outboundConversationEngine = configuredOutboundConversationEngine(options, overrides, conversationEngine, transportMode);
     const needsOpenAIRealtime = conversationEngine === "openai-realtime"
@@ -601,6 +613,7 @@ function safeRuntimeSegment(input) {
     return input.trim().replace(/[^A-Za-z0-9._-]+/g, "-").replace(/^-+|-+$/g, "") || "unknown";
 }
 async function prewarmOutboundGreeting(options, deps) {
+    /* v8 ignore next -- defensive guard against record-play prewarm calls; the implicit-media-stream default added when realtime/SIP credentials are configured prevents this branch from being reachable in current outbound tests @preserve */
     if (options.settings.transportMode !== "media-stream")
         return undefined;
     /* v8 ignore next -- Realtime/SIP outbound tests assert no cascade prewarm is attempted @preserve */

package/dist/senses/voice/twilio-phone.js

@@ -1963,6 +1963,10 @@ class TwilioOpenAIRealtimeMediaStreamSession {
             this.handleCallerSpeechStarted();
             return;
         }
+        if (type === "input_audio_buffer.speech_stopped") {
+            this.handleCallerSpeechStopped();
+            return;
+        }
         if (type === "conversation.item.input_audio_transcription.completed" && typeof event.transcript === "string") {
             this.handleUserTranscript(event.transcript);
             return;
@@ -2032,6 +2036,23 @@ class TwilioOpenAIRealtimeMediaStreamSession {
         clearTimeout(this.pendingUserTurnResponseTimer);
         this.pendingUserTurnResponseTimer = null;
     }
+    handleCallerSpeechStopped() {
+        // VAD signaled the caller stopped speaking. Release the floor immediately
+        // even if transcription has not yet completed (and may never complete for
+        // sub-vocal sounds), so the gate cannot stay stuck thinking the caller
+        // still owns the floor. If a transcript-completed event eventually
+        // arrives, applyCallerTranscriptFinal will run next on an already-released
+        // floor and simply remember the caller turn id.
+        if (!this.activeCallerTurnId)
+            return;
+        if (this.floor.state.floorOwner !== "caller" && this.floor.state.phase !== "caller-speaking")
+            return;
+        this.floor.apply({
+            type: "caller.speech.ended",
+            atMs: Date.now(),
+            turnId: this.activeCallerTurnId,
+        });
+    }
     handleOpenAIAudioDelta(event) {
         const payload = stringField(event.delta);
         if (!payload)
@@ -2070,7 +2091,26 @@ class TwilioOpenAIRealtimeMediaStreamSession {
         const playback = this.playbackState;
         const turnId = `caller-turn-${++this.callerTurnSequence}`;
         this.activeCallerTurnId = turnId;
+        const interruptedResponseId = this.floor.state.activeAssistantSpeechId;
         this.floor.apply({ type: "caller.speech.started", atMs: Date.now(), turnId });
+        if (interruptedResponseId) {
+            // The caller barged in while the floor model still considered an
+            // assistant response active. Without an explicit cancellation, the
+            // assistant.speech.done that eventually arrives leaves floorOwner=caller
+            // permanently set (the reducer only flips owner away from "assistant"
+            // when applying speech.done) and every subsequent caller turn hits the
+            // gate's caller_has_floor block. That is the "Slugger goes silent for
+            // the rest of the call" symptom. Emit a typed cancellation so the floor
+            // model takes the interruption branch and the transcript that follows
+            // can cleanly release the floor.
+            this.floor.apply({
+                type: "assistant.speech.cancelled",
+                atMs: Date.now(),
+                responseId: interruptedResponseId,
+                reason: "caller_barge_in",
+            });
+            this.pendingGatedResponseId = null;
+        }
         this.playbackMarks.clear();
         this.sendTwilioClear();
         if (!playback?.itemId)
@@ -3125,6 +3165,10 @@ class OpenAISipPhoneSession {
             this.clearPendingUserTurnResponse();
             return;
         }
+        if (type === "input_audio_buffer.speech_stopped") {
+            this.handleCallerSpeechStopped();
+            return;
+        }
         if (type === "conversation.item.input_audio_transcription.completed" && typeof event.transcript === "string") {
             this.recordOutboundAmdTranscriptCandidate(event.transcript);
             this.handleUserTranscript(event.transcript);
@@ -3194,6 +3238,23 @@ class OpenAISipPhoneSession {
         clearTimeout(this.pendingUserTurnResponseTimer);
         this.pendingUserTurnResponseTimer = null;
     }
+    handleCallerSpeechStopped() {
+        // VAD signaled the caller stopped speaking. Release the floor immediately
+        // even if transcription has not yet completed (and may never complete for
+        // sub-vocal sounds), so the gate cannot stay stuck thinking the caller
+        // still owns the floor. If a transcript-completed event eventually
+        // arrives, applyCallerTranscriptFinal will run next on an already-released
+        // floor and simply remember the caller turn id.
+        if (!this.activeCallerTurnId)
+            return;
+        if (this.floor.state.floorOwner !== "caller" && this.floor.state.phase !== "caller-speaking")
+            return;
+        this.floor.apply({
+            type: "caller.speech.ended",
+            atMs: Date.now(),
+            turnId: this.activeCallerTurnId,
+        });
+    }
     registerRealtimeToolResponse(responseId, callId) {
         if (!responseId)
             return undefined;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ouro.bot/cli",
-  "version": "0.1.0-alpha.592",
+  "version": "0.1.0-alpha.595",
   "main": "dist/heart/daemon/ouro-entry.js",
   "bin": {
     "cli": "dist/heart/daemon/ouro-bot-entry.js",