From 286683f5db7792be69cc513e9093da28f5e69def Mon Sep 17 00:00:00 2001
From: "rosetta-livekit-bot[bot]"
 <282703043+rosetta-livekit-bot[bot]@users.noreply.github.com>
Date: Thu, 3 Sep 2026 19:53:58 +0000
Subject: [PATCH] feat(assemblyai): surface end-of-turn confidence metadata

---
 .../surface-assemblyai-turn-confidence.md     |  5 +
 plugins/assemblyai/src/stt-metadata.test.ts   | 94 +++++++++++++++++++
 plugins/assemblyai/src/stt.ts                 |  6 ++
 3 files changed, 105 insertions(+)
 create mode 100644 .changeset/surface-assemblyai-turn-confidence.md

diff --git a/.changeset/surface-assemblyai-turn-confidence.md b/.changeset/surface-assemblyai-turn-confidence.md
new file mode 100644
index 0000000000..214ee83bb4
--- /dev/null
+++ b/.changeset/surface-assemblyai-turn-confidence.md
@@ -0,0 +1,5 @@
+---
+'@livekit/agents-plugin-assemblyai': patch
+---
+
+Surface AssemblyAI end-of-turn confidence in speech data metadata.
diff --git a/plugins/assemblyai/src/stt-metadata.test.ts b/plugins/assemblyai/src/stt-metadata.test.ts
index e1fd2e9ab3..2b4ee88bc7 100644
--- a/plugins/assemblyai/src/stt-metadata.test.ts
+++ b/plugins/assemblyai/src/stt-metadata.test.ts
@@ -23,6 +23,7 @@ async function startWebSocketServer() {
 }
 
 async function closeWebSocketServer(wss: WebSocketServer): Promise<void> {
+  for (const client of wss.clients) client.close();
   await new Promise<void>((resolve) => wss.close(() => resolve()));
 }
 
@@ -44,6 +45,53 @@ async function collectUntilEnd(stream: sttLib.SpeechStream): Promise<sttLib.Spee
   return events;
 }
 
+function turnMessage(overrides: Record<string, unknown> = {}): Record<string, unknown> {
+  return {
+    type: 'Turn',
+    words: [{ text: 'hello', start: 0, end: 480, confidence: 0.9 }],
+    end_of_turn: false,
+    transcript: '',
+    ...overrides,
+  };
+}
+
+async function collectTranscript(
+  message: Record<string, unknown>,
+  eventType: sttLib.SpeechEventType,
+): Promise<sttLib.SpeechData> {
+  const { wss, baseUrl } = await startWebSocketServer();
+  let connected = false;
+  let sent = false;
+  const stream = new STT({ apiKey: 'test-key', baseUrl }).stream({
+    connOptions: { maxRetry: 0, retryIntervalMs: 1, timeoutMs: 1000 },
+  });
+
+  wss.on('connection', (ws) => {
+    connected = true;
+    ws.on('message', () => {
+      if (sent) return;
+      sent = true;
+      ws.send(JSON.stringify(message));
+    });
+  });
+
+  try {
+    await waitUntil(() => connected);
+    stream.pushFrame(makeFrame());
+
+    for await (const event of stream) {
+      if (event.type === eventType && event.alternatives?.[0]) {
+        return event.alternatives[0];
+      }
+    }
+
+    throw new Error(`stream ended before ${eventType} was emitted`);
+  } finally {
+    stream.close();
+    await closeWebSocketServer(wss);
+  }
+}
+
 describe('AssemblyAI STT metadata', () => {
   it('maps turn confidence fields onto speech data metadata', async () => {
     const { wss, baseUrl } = await startWebSocketServer();
@@ -129,4 +177,50 @@ describe('AssemblyAI STT metadata', () => {
       await closeWebSocketServer(wss);
     }
   });
+
+  it('surfaces end-of-turn confidence on interim metadata', async () => {
+    const transcript = await collectTranscript(
+      turnMessage({ end_of_turn_confidence: 0.55 }),
+      sttLib.SpeechEventType.INTERIM_TRANSCRIPT,
+    );
+
+    expect(transcript.metadata).toEqual({
+      assemblyai: { endOfTurnConfidence: 0.55 },
+    });
+  });
+
+  it('surfaces end-of-turn confidence on final metadata', async () => {
+    const transcript = await collectTranscript(
+      turnMessage({
+        end_of_turn: true,
+        transcript: 'hello',
+        end_of_turn_confidence: 1,
+      }),
+      sttLib.SpeechEventType.FINAL_TRANSCRIPT,
+    );
+
+    expect(transcript.metadata).toEqual({
+      assemblyai: { endOfTurnConfidence: 1 },
+    });
+  });
+
+  it('surfaces zero end-of-turn confidence', async () => {
+    const transcript = await collectTranscript(
+      turnMessage({ end_of_turn_confidence: 0 }),
+      sttLib.SpeechEventType.INTERIM_TRANSCRIPT,
+    );
+
+    expect(transcript.metadata).toEqual({
+      assemblyai: { endOfTurnConfidence: 0 },
+    });
+  });
+
+  it('leaves metadata unset when end-of-turn confidence is absent', async () => {
+    const transcript = await collectTranscript(
+      turnMessage(),
+      sttLib.SpeechEventType.INTERIM_TRANSCRIPT,
+    );
+
+    expect(transcript.metadata).toBeUndefined();
+  });
 });
diff --git a/plugins/assemblyai/src/stt.ts b/plugins/assemblyai/src/stt.ts
index f2f0db2771..515c40ff4b 100644
--- a/plugins/assemblyai/src/stt.ts
+++ b/plugins/assemblyai/src/stt.ts
@@ -64,6 +64,12 @@ function speechDataMetadata(data: StreamEventMessage): stt.SpeechData['metadata'
     assemblyai.languageConfidence = data.language_confidence;
   }
 
+  // Universal-3 Pro models ramp this from 0 toward 1 while holding a turn open,
+  // allowing callers to trigger eager generation before the final transcript.
+  if (typeof data.end_of_turn_confidence === 'number') {
+    assemblyai.endOfTurnConfidence = data.end_of_turn_confidence;
+  }
+
   if (Object.keys(assemblyai).length === 0) return undefined;
 
   return { assemblyai };