debug: add per-step timing logs to identify latency bottleneck

backend/main.py

@@ -179,22 +179,29 @@ async def conversation_ws(websocket: WebSocket):
                     await websocket.send_json({"type": "error", "message": "No audio data"})
                     continue
 
+                import time
+                t0 = time.time()
                 logger.info(f"[{session_id}] Transcribing {len(audio_buffer)} bytes...")
 
                 # 1. STT
                 transcript = await transcribe_audio(bytes(audio_buffer))
+                t1 = time.time()
                 audio_buffer.clear()
 
                 if not transcript:
                     await websocket.send_json({"type": "error", "message": "Could not transcribe"})
                     continue
 
+                logger.info(f"[{session_id}] STT took {t1-t0:.1f}s")
+
                 await websocket.send_json({"type": "transcript", "role": "user", "text": transcript})
                 conversation_history.append({"role": "user", "content": transcript})
 
                 # 2. LLM (uses cached memory_suffix from identify)
                 logger.info(f"[{session_id}] User: {transcript}")
                 kira_text = await run_conversation(transcript, memory_suffix)
+                t2 = time.time()
+                logger.info(f"[{session_id}] LLM took {t2-t1:.1f}s")
                 conversation_history.append({"role": "assistant", "content": kira_text})
                 logger.info(f"[{session_id}] Kira: {kira_text}")
 
@@ -207,6 +214,8 @@ async def conversation_ws(websocket: WebSocket):
                 # 3. TTS
                 await websocket.send_json({"type": "speaking_start", "text": kira_text})
                 audio_bytes = await synthesize_speech(kira_text)
+                t3 = time.time()
+                logger.info(f"[{session_id}] TTS took {t3-t2:.1f}s. Total: {t3-t0:.1f}s")
                 audio_b64 = base64.b64encode(audio_bytes).decode("utf-8")
                 await websocket.send_json({"type": "audio", "data": audio_b64, "text": kira_text})
                 await websocket.send_json({"type": "speaking_end"})