backend.py

def Backend(pdf_path):

    pdf_name = os.path.basename(pdf_path)

    index_name = os.path.splitext(pdf_name)[0]  # remove extension

    if index_name not in chroma_client.list_collections():
        collection = chroma_client.create_collection(name=index_name)
        # wait a moment for the collection to be fully initialized
        time.sleep(1)

        # First, try to extract text directly from the PDF
        pdf_document = fitz.open(pdf_path)
        texts = []
        for page_number in range(len(pdf_document)):
            page = pdf_document.load_page(page_number)
            texts.append(page.get_text())
        pdf_document.close()

        # If no text is extracted, fall back to the OCR approach
        if not any(texts):
            texts = []

            # Convert PDF to list of images
            images = convert_from_path(pdf_path, output_folder='images')

            # Extract text from each image
            texts = []

            for image in images:
                extracted_text = pytesseract.image_to_string(image)
                texts.append(extracted_text)

        collection.add(documents=texts,
                       metadatas=[{"source": f"page_{i+1}"} for i in range(len(texts))],
                       ids=[f"id{i+1}" for i in range(len(texts))])

    collection = chroma_client.get_collection(index_name)

    responder = Responder(collection)

    return responder