feat(tts): multi-voice switcher for EN & RU, Irina default voice, punctuation pause normalization and speed calibration

app/main.py

@@ -66,6 +66,7 @@ from app.tts import (
     prewarm_card_audio,
     AUDIO_CACHE_DIR,
     is_piper_available,
+    get_available_voices,
 )
 from app.listening import (
     get_listening_exercise,
@@ -523,16 +524,17 @@ async def api_tts(
     text: str = Query(..., min_length=1, max_length=1000),
     lang: str = Query("en"),
     speed: float = Query(1.0, ge=0.5, le=2.0),
+    voice: Optional[str] = Query(None),
     user: dict = Depends(require_user),
 ):
     """
-    Synthesizes speech on-demand using local Piper TTS (en_US-lessac / ru_RU-dmitri).
-    Instantly returns cached audio if previously synthesized.
+    Synthesizes speech on-demand using local Piper TTS.
+    Instantly returns cached audio if previously synthesized for this voice, speed, and text.
     """
     clean_text = text.strip()
     if not clean_text:
         raise HTTPException(status_code=400, detail="Text cannot be empty")
-    audio_path = await synthesize_speech(clean_text, lang=lang, speed=speed)
+    audio_path = await synthesize_speech(clean_text, lang=lang, speed=speed, voice=voice)
     if not audio_path or not audio_path.is_file():
         raise HTTPException(status_code=500, detail="TTS synthesis failed or unavailable")
     return FileResponse(
@@ -544,13 +546,8 @@ async def api_tts(
 
 @app.get("/api/tts/status")
 async def api_tts_status(user: dict = Depends(require_user)):
-    """Returns status of local Piper TTS and available voice models."""
-    return {
-        "available": is_piper_available(),
-        "engine": "piper-vits",
-        "voices": ["en_US-lessac", "ru_RU-dmitri"],
-        "cache_dir": str(AUDIO_CACHE_DIR),
-    }
+    """Returns status of local Piper TTS and all available voice models."""
+    return get_available_voices()
 
 
 # ---------------------------------------------------------------------------
@@ -561,6 +558,7 @@ async def api_tts_status(user: dict = Depends(require_user)):
 async def api_listening_exercise(
     topic: Optional[str] = Query("all"),
     level: Optional[str] = Query("B1"),
+    voice: Optional[str] = Query(None),
     user: dict = Depends(require_user),
 ):
     """
@@ -581,7 +579,7 @@ async def api_listening_exercise(
             ).fetchall()
             user_cards = [dict(r) for r in rows]
 
-    ex = await get_listening_exercise(topic=topic, level=level, user_cards=user_cards)
+    ex = await get_listening_exercise(topic=topic, level=level, user_cards=user_cards, voice=voice)
     # Strip keywords and ideal_answer from client response to prevent inspection
     safe_ex = {k: v for k, v in ex.items() if k not in ("keywords", "ideal_answer")}
     return safe_ex