Stop emitting TTS after speech detection

packages/server/src/server/session.ts

@@ -2072,6 +2072,24 @@ export class Session {
       `[Session ${this.clientId}] Realtime speech chunk detected – aborting playback and LLM`
     );
     this.ttsManager.cancelPendingPlaybacks("realtime speech detected");
+
+    if (this.pendingAudioSegments.length > 0) {
+      console.log(
+        `[Session ${this.clientId}] Dropping ${this.pendingAudioSegments.length} buffered audio segment(s) due to realtime speech`
+      );
+      this.pendingAudioSegments = [];
+    }
+
+    if (this.audioBuffer) {
+      console.log(
+        `[Session ${this.clientId}] Clearing partial audio buffer (${this.audioBuffer.chunks.length} chunk(s))`
+      );
+      this.audioBuffer = null;
+    }
+
+    this.clearBufferTimeout();
+
+    this.abortController.abort();
     await this.handleAbort();
   }
 

plan.md

@@ -60,7 +60,8 @@
   - Added a `speechInProgress` guard in `packages/server/src/server/session.ts` that triggers on the first realtime chunk, cancels pending TTS playback via the new `TTSManager.cancelPendingPlaybacks`, and immediately routes through the abort flow before buffering audio; the flag is cleared when transcription finishes so the next assistant reply can synthesize speech. TTS generation now skips while the flag is set, and `npm run typecheck --workspace=@voice-dev/server` still fails in pre-existing cross-workspace `packages/app/src/types/stream.ts` imports (unchanged by this fix).
 - [x] Interrupt the currently focused agent whenever a realtime user turn begins, ensuring the next prompt starts fresh.
   - Session context now tracks a `focusedAgentId` that each agent screen sets/clears on mount, and the realtime provider looks at that value to send a `cancel_agent_request` before issuing voice aborts so the next spoken prompt doesn't pile onto a running turn. Attempted `npm run typecheck --workspace=@voice-dev/app` but it still fails in the pre-existing stream harness/tests complaining about `parsed*` fields.
-- [ ] Harden playback stop/queue clearing so speech detection purges suppressed audio and the server stops emitting TTS while the user is talking.
+- [x] Harden playback stop/queue clearing so speech detection purges suppressed audio and the server stops emitting TTS while the user is talking.
+  - Speech detection now aborts any in-flight TTS generation, clears buffered realtime segments (pending chunks plus partial buffers), and cancels pending playbacks before delegating to the existing abort flow, so no additional audio_output events fire once the user starts talking.
 - [ ] Prototype chunked audio input: emit smaller PCM frames with `isLast=false`, update `handleAudioChunk` to buffer and trigger STT once the minimum duration is reached.
 - [ ] Investigate streaming STT/server-side VAD (Whisper or equivalent) and document requirements for production use.
 - [ ] Prototype streaming TTS output (chunked `audio_output` messages) and update the player to start playback as soon as chunks arrive.