app.py

from pprint import pprint
from PyPDF2 import PdfReader
import gradio as gr
from transformers import pipeline
import os

# Function to read PDF file content directly
def read_pdf(pdf_path):
 content = ""
 reader = PdfReader(pdf_path)
 for page in reader.pages:
 content += page.extract_text()
 return content

# Process and retrieve answers
def process_invoice(file, questions):
 try:
 # Read the PDF content directly
 print("Reading PDF content...")
 pdf_content = read_pdf(file.name)
 print(f"PDF Content: {pdf_content[:500]}...") # Print first 500 characters for verification

 # Get the Hugging Face token from environment variables
 hf_token = os.getenv("HF_TOKEN")
 if not hf_token:
 raise ValueError("Hugging Face token not found in environment variables.")

 # Initialize the Hugging Face pipeline
 print("Initializing the Hugging Face pipeline...")
 qa_pipeline = pipeline("question-answering", model="mistralai/Mixtral-8x7B-Instruct-v0.1", use_auth_token=hf_token)

 answers = {}
 for question in questions.split(','):
 print(f"Asking question: {question.strip()}")
 result = qa_pipeline(question=question.strip(), context=pdf_content)
 answers[question] = result['answer']
 print(f"Answer: {result['answer']}")

 return answers
 except Exception as e:
 print(f"Error: {e}")
 return {"error": str(e)}

# Gradio interface
def gradio_interface(file, questions):
 answers = process_invoice(file, questions)
 return answers

interface = gr.Interface(
 fn=gradio_interface,
 inputs=[
 gr.File(file_count="single", label="Upload Invoice (PDF)"),
 gr.Textbox(lines=5, placeholder="Enter your questions separated by commas")
 ],
 outputs="json",
 title="Invoice Data Extraction",
 description="Upload an invoice PDF and get the extracted data based on your questions."
)

if __name__ == "__main__":
 interface.launch()