from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse
from pydantic import BaseModel
import sherpa_onnx
import soundfile as sf
import uuid
import os

app = FastAPI(title="K2 Matcha TTS API - Max Quality")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # ganti ke domain frontend lo pas production
    allow_methods=["*"],
    allow_headers=["*"],
)

# Load model Matcha-TTS (kualitas tertinggi sherpa-onnx untuk English)
tts_config = sherpa_onnx.OfflineTtsConfig(
    model=sherpa_onnx.OfflineTtsModelConfig(
        matcha=sherpa_onnx.OfflineTtsMatchaModelConfig(
            acoustic_model="./matcha-icefall-en_US-ljspeech/model-steps-3.onnx",
            vocoder="./vocos-22khz-univ.onnx",
            tokens="./matcha-icefall-en_US-ljspeech/tokens.txt",
            data_dir="./matcha-icefall-en_US-ljspeech/espeak-ng-data",
        ),
        num_threads=2,
    )
)
tts = sherpa_onnx.OfflineTts(tts_config)

os.makedirs("outputs", exist_ok=True)

class TTSRequest(BaseModel):
    text: str
    speed: float = 1.0

@app.get("/")
def root():
    return {"status": "K2 Matcha TTS API is running", "quality": "max"}

@app.post("/tts")
def generate_tts(req: TTSRequest):
    audio = tts.generate(req.text, sid=0, speed=req.speed)
    filename = f"outputs/{uuid.uuid4().hex}.wav"
    sf.write(filename, audio.samples, audio.sample_rate)
    return FileResponse(filename, media_type="audio/wav", filename="output.wav")