73 lines
1.9 KiB
Python
73 lines
1.9 KiB
Python
"""ASR-микросервис: whisper-large-v3-russian + Qwen3-ASR-1.7B для calls24."""
|
|
import logging
|
|
|
|
from fastapi import FastAPI, UploadFile
|
|
from fastapi.responses import JSONResponse
|
|
from pydantic import BaseModel
|
|
from starlette.middleware.cors import CORSMiddleware
|
|
|
|
from .config import settings
|
|
from .engine import load_models
|
|
|
|
logging.basicConfig(level=logging.INFO)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
app = FastAPI(title="ASR Microservice", version="1.0.0")
|
|
|
|
app.add_middleware(
|
|
CORSMiddleware,
|
|
allow_origins=["*"],
|
|
allow_methods=["*"],
|
|
allow_headers=["*"],
|
|
)
|
|
|
|
_models = {}
|
|
|
|
|
|
@app.on_event("startup")
|
|
async def startup_event():
|
|
import torch
|
|
|
|
device = torch.device(settings.DEVICE)
|
|
global _models
|
|
_models = load_models(device)
|
|
logger.info(f"Модели загружены: {list(_models.keys())}")
|
|
|
|
|
|
@app.get("/health")
|
|
async def health():
|
|
return {"status": "ok", "models": list(_models.keys())}
|
|
|
|
|
|
@app.post("/transcribe")
|
|
async def transcribe(file: UploadFile, model: str = "qwen3"):
|
|
"""Распознать загруженный аудиофайл (model в query-параметре)."""
|
|
import os
|
|
import tempfile
|
|
|
|
tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
|
|
content = await file.read()
|
|
tmp.write(content)
|
|
tmp.close()
|
|
|
|
if model not in _models:
|
|
return JSONResponse(
|
|
status_code=400,
|
|
content={"error": f"Неизвестная модель: {model}. Доступны: {list(_models.keys())}"},
|
|
)
|
|
|
|
try:
|
|
result = _models[model].transcribe(tmp.name)
|
|
return {"model": model, "text": result}
|
|
except Exception as e:
|
|
logger.error(f"Ошибка распознавания: {e}")
|
|
return JSONResponse(status_code=500, content={"error": str(e)})
|
|
finally:
|
|
os.unlink(tmp.name)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
import uvicorn
|
|
|
|
uvicorn.run(app, host="0.0.0.0", port=8000)
|