app.py

from flask import Flask, request, jsonify, send_from_directory
import requests
import os
from dotenv import load_dotenv
import logging

load_dotenv()

app = Flask(__name__)

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

API_URL = "/static-proxy?url=https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2F%26quot%3B%3C%2Fspan%3E%3C!-- HTML_TAG_END -->
headers = {"Authorization": f"Bearer {os.getenv('HUGGINGFACE_API_KEY')}"}
HF_REPO_ID = os.getenv('HF_REPO_ID')
PDF_FILENAME = os.getenv('PDF_FILENAME')

def query(payload, model):
    try:
        response = requests.post(API_URL + model, headers=headers, json=payload)
        response.raise_for_status()
        return response.json()
    except requests.exceptions.RequestException as e:
        logger.error(f"Error querying model {model}: {str(e)}")
        return {"error": str(e)}

def process_pdf():
    pdf_url = f"https://huggingface.co/spaces/{HF_REPO_ID}/resolve/main/data/{PDF_FILENAME}"
    logger.info(f"Attempting to process PDF at URL: {pdf_url}")
    
    model = "impira/layoutlm-document-qa"
    payload = {
        "inputs": {
            "question": "Summarize the main points of this document.",
            "image": pdf_url
        }
    }
    
    response = query(payload, model)
    logger.info(f"PDF processing response: {response}")
    
    if 'error' in response:
        return f"Error processing PDF: {response['error']}"
    elif 'answer' in response:
        return response['answer']
    else:
        return str(response)

# Process PDF and get summary
pdf_summary = process_pdf()
logger.info(f"PDF Summary: {pdf_summary}")

# Get embeddings for the summary
embedding_model = "sentence-transformers/all-MiniLM-L6-v2"
summary_embedding = query({"inputs": pdf_summary}, embedding_model)[0] if not pdf_summary.startswith("Error") else None

@app.route('/')
def home():
    return send_from_directory('.', 'index.html')

@app.route('/ask', methods=['POST'])
def ask():
    prompt = request.json['question']
    
    if summary_embedding is None:
        return jsonify({'response': "I'm sorry, but I couldn't process the PDF correctly. Please check the PDF file and try again later."})
    
    # Get embedding for the question
    query_embedding = query({"inputs": prompt}, embedding_model)[0]
    
    # Calculate similarity
    similarity = sum(a*b for a, b in zip(query_embedding, summary_embedding))
    
    # Generate response using T5 model
    generator_model = "google/flan-t5-base"
    input_text = f"Context: {pdf_summary}\n\nQuestion: {prompt}\n\nAnswer:"
    response = query({"inputs": input_text}, generator_model)[0]["generated_text"]

    return jsonify({'response': response})

if __name__ == '__main__':
    app.run(debug=True)