app.py

import fitz  # PyMuPDF
import os
import pytesseract
from PIL import Image
import io
from llama_index import *
from transformers import pipeline
from fastapi import FastAPI, UploadFile, File, Form

app = FastAPI()

# Endpoint to check for files in the current directory
@app.get("/check_files")
async def check_files():
    files = os.listdir(".")
    if any(file.endswith(".pdf") for file in files):
        return {"message": "Files found in directory. You can proceed."}
    else:
        return {"message": "Select a file to work on in PDF."}

# Function to extract text from PDF with OCR
def extract_text_from_pdf(pdf_path):
    doc = fitz.open(pdf_path)
    text = ""
    
    for page_num in range(len(doc)):
        page = doc.load_page(page_num)
        
        # Extract text from the page
        text += page.get_text()
        
        # Extract images from the page
        image_list = page.get_images(full=True)
        
        for img_index, img in enumerate(image_list):
            xref = img[0]
            base_image = doc.extract_image(xref)
            image_bytes = base_image["image"]
            image_ext = base_image["ext"]
            
            # Load image
            image = Image.open(io.BytesIO(image_bytes))
            
            # Use Tesseract to do OCR on the image
            image_text = pytesseract.image_to_string(image)
            
            # Append OCR text to the main text
            text += image_text

    return text

# Function to create the index
def create_index(text):
    # Save text to a file
    with open("pdf_text.txt", "w") as file:
        file.write(text)

    reader = SimpleDirectoryReader(".")
    index = GPTVectorStoreIndex.from_documents(reader.load_data())
    return index

# Function to answer a question
def answer_question(question, index, qa_pipeline):
    # Retrieve the most relevant section from the index
    response = index.query(question)
    
    # Use the Hugging Face QA pipeline to find the answer in the retrieved text
    result = qa_pipeline(question=question, context=response)
    
    return result['answer']

# Use a Hugging Face model for question answering
qa_pipeline = pipeline("question-answering", model="distilbert-base-uncased-distilled-squad")

@app.post("/upload")
async def upload_file(file: UploadFile = File(...), question: str = Form(...)):
    pdf_text = extract_text_from_pdf(file.file)
    index = create_index(pdf_text)
    answer = answer_question(question, index, qa_pipeline)
    return {"answer": answer}