"""FastAPI app for Hugging Face Spaces — lightweight Sauti TTS.

Deploy as HF Space (CPU, Free).
"""

import os
import uuid
import logging
import time
from pathlib import Path

import numpy as np
import soundfile as sf
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

app = FastAPI(title="Sauti TTS Lightweight")

# Global inference engine
engine = None


class TTSRequest(BaseModel):
    text: str
    ref_text: str = ""
    steps: int = 10
    cfg: float = 1.5
    speed: float = 1.0
    seed: int | None = None


@app.on_event("startup")
def startup():
    global engine
    from lightweight_infer import LightweightSautiInference

    logger.info("Loading model...")
    t0 = time.time()
    engine = LightweightSautiInference(
        checkpoint=os.getenv("HF_MODEL_ID", "msingiai/sauti-tts"),
        vocab=os.getenv("HF_MODEL_ID", "msingiai/sauti-tts"),
        device="cpu",
        quantize=True,
        nfe_steps=10,
        cfg_strength=1.5,
    )
    logger.info(f"Model loaded in {time.time() - t0:.1f}s")


@app.get("/health")
def health():
    return {"status": "ok", "model_loaded": engine is not None}


@app.post("/tts")
def tts(req: TTSRequest):
    if engine is None:
        raise HTTPException(status_code=503, detail="Model not loaded")

    # Use a default reference audio if none provided
    ref = os.getenv("DEFAULT_REF_AUDIO", "")
    if not ref:
        raise HTTPException(status_code=400, detail="No reference audio configured")

    t0 = time.time()
    try:
        audio, sr = engine.generate(
            text=req.text,
            ref_audio_path=ref,
            ref_text=req.ref_text,
            speed=req.speed,
            seed=req.seed,
        )
    except Exception as e:
        logger.exception("Inference failed")
        raise HTTPException(status_code=500, detail=str(e))

    elapsed = time.time() - t0
    out_path = Path("/tmp") / f"{uuid.uuid4().hex}.wav"
    sf.write(str(out_path), audio, sr)

    rtf = elapsed / (len(audio) / sr)
    return {
        "audio_path": str(out_path),
        "sample_rate": sr,
        "duration": len(audio) / sr,
        "inference_sec": elapsed,
        "rtf": rtf,
    }