app.py

from fastapi import FastAPI, Request, HTTPException
from pydantic import BaseModel
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import requests
import logging
from contextlib import asynccontextmanager

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Global variables for model and tokenizer
model = None
tokenizer = None
label_mapping = None

@asynccontextmanager
async def lifespan(app: FastAPI):
    """Load the model and tokenizer on startup"""
    global model, tokenizer, label_mapping
    
    try:
        model_name = "ityndall/james-river-classifier"
        logger.info(f"Loading model: {model_name}")
        
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForSequenceClassification.from_pretrained(model_name)
        
        # Load label mapping
        label_mapping_url = f"https://huggingface.co/{model_name}/resolve/main/label_mapping.json"
        response = requests.get(label_mapping_url)
        label_mapping = response.json()
        
        logger.info("Model loaded successfully")
        logger.info(f"Available labels: {list(label_mapping['id2label'].values())}")
        
    except Exception as e:
        logger.error(f"Error loading model: {str(e)}")
        raise e
    
    yield
    
    # Cleanup (if needed)
    logger.info("Shutting down...")

app = FastAPI(
    title="James River Survey Classification API",
    description="API for classifying survey-related text messages into job types",
    version="1.0.0",
    lifespan=lifespan
)

# Request model
class PredictionRequest(BaseModel):
    message: str

# Response model
class PredictionResponse(BaseModel):
    label: str
    confidence: float

@app.get("/")
async def root():
    """Root endpoint with API information"""
    return {
        "message": "James River Survey Classification API",
        "version": "1.0.0",
        "model": "ityndall/james-river-classifier",
        "available_labels": list(label_mapping["id2label"].values()) if label_mapping else [],
        "endpoints": {
            "predict": "/predict - POST endpoint for text classification",
            "health": "/health - GET endpoint for health check"
        }
    }

@app.get("/health")
async def health_check():
    """Health check endpoint"""
    if model is None or tokenizer is None or label_mapping is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    return {"status": "healthy", "model_loaded": True}

@app.post("/predict", response_model=PredictionResponse)
async def predict(request: PredictionRequest):
    """Predict the survey job type for the given message"""
    
    if model is None or tokenizer is None or label_mapping is None:
        raise HTTPException(status_code=503, detail="Model not loaded")
    
    try:
        text = request.message.strip()
        if not text:
            raise HTTPException(status_code=400, detail="Message cannot be empty")
        
        # Tokenize and predict
        inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128)
        
        with torch.no_grad():
            logits = model(**inputs).logits
            probs = torch.nn.functional.softmax(logits, dim=-1)
            predicted_class_id = probs.argmax().item()
            confidence = probs[0][predicted_class_id].item()
        
        # Get label
        label = label_mapping["id2label"][str(predicted_class_id)]
        
        logger.info(f"Prediction: '{text}' -> {label} (confidence: {confidence:.3f})")
        
        return PredictionResponse(label=label, confidence=confidence)
        
    except Exception as e:
        logger.error(f"Error during prediction: {str(e)}")
        raise HTTPException(status_code=500, detail=f"Prediction error: {str(e)}")

# Legacy endpoint for backward compatibility
@app.post("/predict_legacy")
async def predict_legacy(request: Request):
    """Legacy endpoint that accepts raw JSON (for backward compatibility)"""
    try:
        data = await request.json()
        message = data.get("message", "")
        
        if not message:
            raise HTTPException(status_code=400, detail="Message field is required")
        
        # Use the main predict function
        prediction_request = PredictionRequest(message=message)
        result = await predict(prediction_request)
        
        return {"label": result.label, "confidence": result.confidence}
        
    except Exception as e:
        logger.error(f"Error in legacy endpoint: {str(e)}")
        raise HTTPException(status_code=500, detail=str(e))

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