app.py

from fastapi import FastAPI
from fastapi.responses import FileResponse
from fastapi.staticfiles import StaticFiles
from fastapi.middleware.cors import CORSMiddleware
from transformers import pipeline
import os
import uvicorn
import logging

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

# Set cache directory to a writable location
cache_dir = "/tmp/hf_cache"
os.environ["HF_HOME"] = cache_dir
os.environ["HUGGINGFACE_HUB_CACHE"] = cache_dir
os.environ["TRANSFORMERS_CACHE"] = cache_dir

# Create the cache directory if it doesn't exist
if not os.path.exists(cache_dir):
    os.makedirs(cache_dir, exist_ok=True)

app = FastAPI()

# Add CORS middleware to allow frontend requests
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

app.mount("/static", StaticFiles(directory="static"), name="static")

# Load the zero-shot classification model with explicit cache directory
logger.info("Loading the model...")
try:
    classifier = pipeline(
        "zero-shot-classification",
        model="UBC-NLP/ARBERTv2",  # Switch to a better Arabic model
        tokenizer="UBC-NLP/ARBERTv2",
        cache_dir=cache_dir
    )
    logger.info("Model loaded successfully!")
except Exception as e:
    logger.error(f"Error loading model: {str(e)}")
    raise

@app.get("/")
async def index():
    logger.info("Serving index.html")
    return FileResponse("static/index.html")

@app.post("/classify")
async def classify_text(data: dict):
    logger.info(f"Received classify request with data: {data}")
    try:
        text = data.get("document")
        labels = data.get("labels")
        if not text or not labels:
            logger.warning("Missing text or labels in request")
            return {"error": "Please provide both text and labels"}, 400
        
        # Convert labels to list if it's a string
        if isinstance(labels, str):
            labels = [label.strip() for label in labels.split(",") if label.strip()]
        
        logger.info(f"Classifying text: {text[:50]}... with labels: {labels}")
        result = classifier(text, labels, multi_label=False)
        logger.info(f"Classification result: {result}")
        return {"labels": result["labels"], "scores": result["scores"]}
    except Exception as e:
        logger.error(f"Error during classification: {str(e)}")
        return {"error": str(e)}, 500

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