app.py

from fastapi import FastAPI
from pydantic import BaseModel
from contextlib import asynccontextmanager
import os

# Global variables
model = None
tokenizer = None
device = None
model_loaded = False

def load_model():
    global model, tokenizer, device, model_loaded
    try:
        print("🔄 Starting model loading...")
        
        # Set cache directory
        os.environ["HF_HOME"] = "/tmp"
        
        # Import here to avoid startup issues
        from transformers import T5ForConditionalGeneration, T5Tokenizer
        import torch
        
        print("📦 Loading tokenizer...")
        tokenizer = T5Tokenizer.from_pretrained("chalana2001/quiz_guru_chatbot")
        
        print("🤖 Loading model...")
        model = T5ForConditionalGeneration.from_pretrained(
            "chalana2001/quiz_guru_chatbot", 
            trust_remote_code=True
        )
        
        device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
        model.to(device)
        model_loaded = True
        
        print(f"✅ Model loaded successfully on {device}")
        return True
        
    except Exception as e:
        print(f"❌ Error loading model: {e}")
        return False

@asynccontextmanager
async def lifespan(app: FastAPI):
    # Startup
    print("🚀 Starting up...")
    load_model()
    yield
    # Shutdown (if needed)
    print("🛑 Shutting down...")

app = FastAPI(title="Quiz Guru Chatbot", version="1.0.0", lifespan=lifespan)

class PromptRequest(BaseModel):
    prompt: str

@app.get("/")
def read_root():
    return {
        "message": "Quiz Guru Chatbot API", 
        "status": "running",
        "model_loaded": model_loaded
    }

@app.get("/health")
def health():
    return {
        "status": "healthy",
        "model_loaded": model_loaded,
        "device": str(device) if device else "unknown"
    }

@app.post("/predict")
def predict(request: PromptRequest):
    if not model_loaded:
        return {"error": "Model not loaded. Please check /health endpoint."}
    
    try:
        # Import torch here
        import torch
        
        inputs = tokenizer(request.prompt, return_tensors="pt", padding=True).to(device)
        
        with torch.no_grad():
            output = model.generate(**inputs, max_length=256, num_beams=4, early_stopping=True)
        
        decoded = tokenizer.decode(output[0], skip_special_tokens=True)
        
        return {"result": decoded, "status": "success"}
        
    except Exception as e:
        return {"error": str(e), "status": "error"}

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)