app.py

from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
import os
from typing import Optional

# キャッシュディレクトリの設定
os.environ["HF_HOME"] = "/app/.cache"
os.environ["TRANSFORMERS_CACHE"] = "/app/.cache"
os.environ["HUGGINGFACE_HUB_CACHE"] = "/app/.cache"

# キャッシュディレクトリを作成
cache_dir = "/app/.cache"
os.makedirs(cache_dir, exist_ok=True)

app = FastAPI(title="Lightweight Hugging Face API")

class TextRequest(BaseModel):
    text: str
    max_length: Optional[int] = 50

class SentimentResponse(BaseModel):
    text: str
    sentiment: str
    confidence: float
    model_name: str

class GenerateResponse(BaseModel):
    input_text: str
    generated_text: str
    model_name: str

# グローバル変数
sentiment_classifier = None
text_generator = None

@app.on_event("startup")
async def load_models():
    """軽量モデルをロード"""
    global sentiment_classifier, text_generator
    
    print("🚀 軽量モデルのロード開始...")
    
    try:
        from transformers import pipeline
        
        # 非常に軽量な感情分析モデル（cache_dirを削除）
        print("📥 軽量感情分析モデルをロード中...")
        sentiment_classifier = pipeline(
            "sentiment-analysis",
            model="prajjwal1/bert-tiny"
        )
        print("✅ 感情分析モデル読み込み完了")
        
        # 軽量テキスト生成モデル（cache_dirを削除）
        print("📥 軽量テキスト生成モデルをロード中...")
        text_generator = pipeline(
            "text-generation",
            model="sshleifer/tiny-gpt2"
        )
        print("✅ テキスト生成モデル読み込み完了")
        
        print("✅ 全てのモデル読み込み完了")
        
    except Exception as e:
        print(f"❌ モデル読み込みエラー: {e}")
        import traceback
        traceback.print_exc()

@app.get("/")
async def root():
    return {
        "message": "🤗 Lightweight Hugging Face API is running!",
        "status": "healthy",
        "models": "lightweight versions"
    }

@app.post("/sentiment", response_model=SentimentResponse)
async def analyze_sentiment(request: TextRequest):
    """軽量感情分析"""
    try:
        if sentiment_classifier is None:
            raise HTTPException(status_code=503, detail="Sentiment model not loaded")
        
        result = sentiment_classifier(request.text)
        
        return SentimentResponse(
            text=request.text,
            sentiment=result[0]["label"],
            confidence=round(result[0]["score"], 4),
            model_name="prajjwal1/bert-tiny"
        )
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Analysis failed: {str(e)}")

@app.post("/generate", response_model=GenerateResponse)
async def generate_text(request: TextRequest):
    """軽量テキスト生成"""
    try:
        if text_generator is None:
            raise HTTPException(status_code=503, detail="Generation model not loaded")
        
        max_length = min(request.max_length, 100)
        
        result = text_generator(
            request.text,
            max_length=max_length,
            num_return_sequences=1,
            temperature=0.7,
            do_sample=True,
            pad_token_id=text_generator.tokenizer.eos_token_id
        )
        
        return GenerateResponse(
            input_text=request.text,
            generated_text=result[0]["generated_text"],
            model_name="sshleifer/tiny-gpt2"
        )
    
    except Exception as e:
        raise HTTPException(status_code=500, detail=f"Generation failed: {str(e)}")

@app.get("/models")
async def get_models():
    return {
        "sentiment_analysis": {
            "model": "prajjwal1/bert-tiny",
            "status": "loaded" if sentiment_classifier else "not loaded"
        },
        "text_generation": {
            "model": "sshleifer/tiny-gpt2", 
            "status": "loaded" if text_generator else "not loaded"
        },
        "note": "Using lightweight models for Spaces compatibility"
    }

if __name__ == "__main__":
    import uvicorn
    port = int(os.environ.get("PORT", 7860))
    uvicorn.run(app, host="0.0.0.0", port=port)