main.py

#main.py
from fastapi import FastAPI, HTTPException, UploadFile, File
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import RedirectResponse
from fastapi.staticfiles import StaticFiles
from pydantic import BaseModel
from typing import List
import numpy as np
import pandas as pd
import base64
import io
import datetime

from app.utils import trim_zero_padding, sliding_window_gesture_detection
from app.model_loader import load_models
from app.gpt_router import router as gpt_router
from app.converter import convert_gestures_to_sentence
from app.tts import generate_tts

app = FastAPI()

# 정적 파일 제공
app.mount("/tts_audios", StaticFiles(directory="tts_audios"), name="tts")

@app.get("/", include_in_schema=False)
async def root():
    return RedirectResponse(url="/docs")

# CORS 허용
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_methods=["*"],
    allow_headers=["*"],
)

app.include_router(gpt_router)

# 모델 로딩
encoder_model, gesture_hmms, ergodic_model = load_models()

# 응답 스키마 정의
class Interval(BaseModel):
    start: int
    end: int
    label: str

class GestureResponse(BaseModel):
    intervals: List[Interval]

class TranslateResponse(BaseModel):
    sentence: str
    audio_base64: str

class TTSRequest(BaseModel):
    sentence: str

# CSV 파일로 제스처 인식
@app.post("/predict_gesture_from_csv", response_model=GestureResponse)
async def predict_gesture_from_csv(file: UploadFile = File(...)):
    try:
        contents = await file.read()
        df = pd.read_csv(io.BytesIO(contents))
        data = df.values.astype(np.float32)

        trimmed = trim_zero_padding(data)

        t1 = datetime.datetime.now()
        intervals = sliding_window_gesture_detection(
            continuous_sequence=trimmed,
            encoder_model=encoder_model,
            gesture_hmms=gesture_hmms,
            final_model=ergodic_model,
            window_size=10,
            step=2,
            threshold_diff=0.0,
            min_merge_gap=3,
            min_interval_length=3
        )
        t2 = datetime.datetime.now()
        print("수화인식 결과 반환api",t2-t1)

        return GestureResponse(
            intervals=[Interval(start=int(s), end=int(e), label=l) for s, e, l in intervals]
        )

    except Exception as e:
        raise HTTPException(status_code=400, detail=str(e))

# CSV 파일로 제스처 → 문장
@app.post("/predict_gesture_and_sentence_from_csv", summary="CSV 기반 제스처를 문장으로 변환", tags=["Gesture → Sentence"])
async def predict_gesture_and_sentence_from_csv(file: UploadFile = File(...)):
    try:
        contents = await file.read()
        df = pd.read_csv(io.BytesIO(contents))
        data = df.values.astype(np.float32)

        trimmed = trim_zero_padding(data)

        t1 = datetime.datetime.now()
        intervals = sliding_window_gesture_detection(
            continuous_sequence=trimmed,
            encoder_model=encoder_model,
            gesture_hmms=gesture_hmms,
            final_model=ergodic_model,
            window_size=10,
            step=2,
            threshold_diff=0.0,
            min_merge_gap=3,
            min_interval_length=3
        )
        t2 = datetime.datetime.now()
        print("결과",t2-t1)

        gestures = [label for _, _, label in intervals] or ["none"]
        sentence = convert_gestures_to_sentence(gestures)

        return {"sentence": sentence}

    except Exception as e:
        raise HTTPException(status_code=400, detail=str(e))

# TTS
@app.post("/generate_tts", response_model=TranslateResponse, summary="자연어 문장을 TTS 음성(base64)으로 변환", tags=["TTS"])
async def generate_tts_audio(data: TTSRequest):
    try:
        sentence = data.sentence
        audio_path = generate_tts(sentence)

        with open(audio_path, "rb") as f:
            audio_base64 = base64.b64encode(f.read()).decode("utf-8")

        return {
            "sentence": sentence,
            "audio_base64": audio_base64
        }

    except Exception as e:
        raise HTTPException(status_code=400, detail=str(e))

# 실행
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)