mirror of
https://github.com/R0m1k3/Socialflow.git
synced 2026-10-11 17:26:45 +02:00
Claude
fdace9d8d5
- qwen-tts : GPU NVIDIA détecté automatiquement (bfloat16), image CUDA 12.8 (RTX 50xx comprises) via TORCH_VARIANT=cu128, docker-compose.gpu.yml dédié (clé obligatoire, choix de la carte) et guide d'installation Unraid. - Paramètres → « Qwen TTS (voix locale) » : adresse + clé du service, test de connexion (modèle et carte graphique affichés), stockés en base (app_config.qwen_tts_url / qwen_tts_api_key) et transmis à ffmpeg-api à chaque voix ; QWEN_TTS_URL reste un défaut. - Le service qwen-tts CPU du docker-compose principal devient optionnel (profil qwen-local). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Upu97wMmsRkBoj6iVM4rH6
76 lines
2.9 KiB
Python
76 lines
2.9 KiB
Python
"""Synthèse vocale Qwen3-TTS (service qwen-tts, local ou sur un serveur GPU)."""
|
|
|
|
import logging
|
|
from dataclasses import dataclass
|
|
from pathlib import Path
|
|
|
|
import httpx
|
|
|
|
from .. import config
|
|
from ..voices import qwen_instruction, resolve_qwen_voice
|
|
|
|
log = logging.getLogger(__name__)
|
|
|
|
|
|
class QwenError(RuntimeError):
|
|
pass
|
|
|
|
|
|
@dataclass(frozen=True)
|
|
class Target:
|
|
"""Service Qwen à appeler : réglé dans les paramètres de l'application,
|
|
sinon par les variables d'environnement."""
|
|
|
|
url: str
|
|
api_key: str = ""
|
|
|
|
@classmethod
|
|
def resolve(cls, url: str | None = None, api_key: str | None = None) -> "Target | None":
|
|
if url and url.strip():
|
|
return cls(url.strip().rstrip("/"), (api_key or "").strip())
|
|
if config.QWEN_TTS_URL:
|
|
return cls(config.QWEN_TTS_URL, config.QWEN_TTS_API_KEY)
|
|
return None
|
|
|
|
@property
|
|
def headers(self) -> dict[str, str]:
|
|
return {"x-api-key": self.api_key} if self.api_key else {}
|
|
|
|
|
|
async def synthesize(
|
|
text: str, voice: str | None, style: str | None, target: Target | None, workdir: Path, attempt: int = 0
|
|
) -> tuple[Path, str]:
|
|
"""Génère la voix ; renvoie un WAV brut et la voix utilisée."""
|
|
if target is None:
|
|
raise QwenError("service Qwen TTS non configuré (Paramètres → Qwen TTS)")
|
|
qwen_voice = resolve_qwen_voice(voice)
|
|
payload = {"text": text, "voice": qwen_voice, "instruct": qwen_instruction(style), "language": "French"}
|
|
log.info("Qwen TTS (%s) : voix=%s style=%s (%d car.)", target.url, qwen_voice, style, len(text))
|
|
try:
|
|
async with httpx.AsyncClient(timeout=config.QWEN_TTS_TIMEOUT) as client:
|
|
response = await client.post(f"{target.url}/synthesize", json=payload, headers=target.headers)
|
|
except httpx.HTTPError as error:
|
|
raise QwenError(f"service Qwen TTS injoignable : {error!r}") from error
|
|
if response.status_code != 200:
|
|
raise QwenError(f"Qwen TTS : HTTP {response.status_code} {response.text[:300]}")
|
|
raw = workdir / f"qwen_raw_{attempt}.wav"
|
|
raw.write_bytes(response.content)
|
|
return raw, response.headers.get("x-voice", qwen_voice)
|
|
|
|
|
|
async def status(target: Target | None) -> dict | None:
|
|
"""État du service (modèle, carte graphique) et voix proposées, clonées
|
|
comprises ; None s'il n'est pas configuré ou ne répond pas."""
|
|
if target is None:
|
|
return None
|
|
try:
|
|
async with httpx.AsyncClient(timeout=10) as client:
|
|
health = await client.get(f"{target.url}/health", headers=target.headers)
|
|
health.raise_for_status()
|
|
listed = await client.get(f"{target.url}/voices", headers=target.headers)
|
|
listed.raise_for_status()
|
|
return {**health.json(), "voices": listed.json()["voices"]}
|
|
except Exception as error: # noqa: BLE001 — moteur simplement signalé indisponible
|
|
log.warning("Service Qwen TTS %s indisponible : %s", target.url, error)
|
|
return None
|