perf(bridge): lock STT to Korean + add per-stage turn timing

- transcribe() now passes language="ko" (STT_LANGUAGE env, default ko): skips
  Whisper auto-detect, fixing occasional Korean->Chinese mis-detection and
  shaving latency. LLM is already locked via OUTPUT_LANGUAGE=Korean; MeloTTS is
  Korean-only — so STT/LLM/TTS are all Korean now.
- converse_stream logs "⏱️ turn stt=.. think(LLM)=.. tts=.. total=.." so the
  ~30s voice-reply latency can be attributed to the real bottleneck stage.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

bridge/server.py

@@ -75,6 +75,11 @@ VAD_ENABLED = os.environ.get("VAD_ENABLED", "1") not in ("0", "false", "False")
 VAD_THRESHOLD = float(os.environ.get("VAD_THRESHOLD", "0.4"))
 VAD_MIN_SPEECH_MS = int(os.environ.get("VAD_MIN_SPEECH_MS", "200"))
 
+# Lock STT to a single language (this deployment is Korean-only). Skipping
+# Whisper's language auto-detect both fixes occasional mis-detection (e.g. a
+# Korean phrase decoded as Chinese) and shaves a little latency. Empty = auto.
+STT_LANGUAGE = os.environ.get("STT_LANGUAGE", "ko").strip() or None
+
 # TTS engine: "melo" (MeloTTS Korean speaker, the warm worker) is the primary
 # voice; Piper is kept as a fallback if the worker is unreachable. Set
 # TTS_ENGINE=piper to disable MeloTTS entirely.
@@ -208,7 +213,7 @@ def transcribe(wav_bytes: bytes) -> dict:
         print("[bridge] no speech detected (VAD) — skipping STT", flush=True)
         return {"text": "", "language": None}
 
-    segments, info = _whisper.transcribe(audio, beam_size=1)
+    segments, info = _whisper.transcribe(audio, beam_size=1, language=STT_LANGUAGE)
     # Second line of defence: drop non-speech / hallucinated segments by
     # Whisper's own no_speech_prob. The no_speech_prob hard cutoff (plus the VAD
     # pre-gate above) is what rejects noise/hallucinations. The avg_logprob
@@ -446,7 +451,10 @@ def http_converse_stream():
     broadcasting = _coerce_bool(request.args.get("broadcasting"))
 
     def gen():
+        import time
+        t0 = time.monotonic()
         stt = transcribe(raw)
+        t_stt = time.monotonic()
         transcript = stt.get("text", "")
         if not transcript:
             yield json.dumps({"type": "meta", "transcript": "", "language": stt.get("language"),
@@ -454,6 +462,7 @@ def http_converse_stream():
             yield json.dumps({"type": "end"}) + "\n"
             return
         result = think(transcript, stt.get("language"), broadcasting)
+        t_think = time.monotonic()
         reply = result.get("reply", "")
         yield json.dumps({
             "type": "meta",
@@ -463,8 +472,11 @@ def http_converse_stream():
             "error": result.get("error"),
             "broadcast_action": result.get("broadcast_action"),
         }) + "\n"
+        tts_total = 0.0
         for seq, sentence in enumerate(split_sentences(reply)):
+            ts = time.monotonic()
             audio = synthesize(sentence)
+            tts_total += time.monotonic() - ts
             if audio:
                 yield json.dumps({
                     "type": "audio",
@@ -472,6 +484,12 @@ def http_converse_stream():
                     "audio_b64": base64.b64encode(audio).decode("ascii"),
                 }) + "\n"
         yield json.dumps({"type": "end"}) + "\n"
+        print(
+            f"[bridge] ⏱️ turn stt={t_stt - t0:.1f}s think(LLM)={t_think - t_stt:.1f}s "
+            f"tts={tts_total:.1f}s total={time.monotonic() - t0:.1f}s replylen={len(reply)} "
+            f"transcript={transcript[:40]!r}",
+            flush=True,
+        )
 
     return Response(stream_with_context(gen()), mimetype="application/x-ndjson")