PDF Information and Inference 🗞️

Overview

This project provides a Gradio interface to upload a PDF, extract its content, and interact with it using natural language queries. The model answers questions based on the content of the uploaded PDF, using a combination of HuggingFace and Llama Index.

Features

• Upload PDF files.
• Extract and display PDF content.
• Ask questions about the PDF content.
• Get answers from the content using a Q&A assistant.

Setup

Prerequisites

• Python 3.8+
• Install the necessary libraries using pip: bash pip install gradio llama-index huggingface_hub python-dotenv langdetect

Environment Setup

1. HuggingFace Token: Replace args with your HuggingFace token: python args = "your_huggingface_token"
2. Directory Setup: Ensure the directories for data storage exist: python PERSIST_DIR = "./db" DATA_DIR = "data"

Running the Application

1. Login to HuggingFace: python command = f"huggingface-cli login --token {args}" subprocess.run(command, shell=True, check=True)

2. Configure Llama Index:

   • Set up the language model and embedding model to use GPU if available: python Settings.llm = HuggingFaceInferenceAPI( model_name="mistralai/Mistral-7B-Instruct-v0.3", tokenizer_name="mistralai/Mistral-7B-Instruct-v0.3", context_window=5000, max_new_tokens=1024, generate_kwargs={"temperature": 0.1}, device=0 # Use GPU 0, set to -1 for CPU ) Settings.embed_model = HuggingFaceEmbedding( model_name="BAAI/bge-small-en-v1.5", device=0 # Use GPU 0, set to -1 for CPU )

3. Launch the Gradio Interface: python demo.launch()

Usage

1. Upload PDF:

   • Click "Upload your PDF Files" and select your PDF.
   • Click "Upload PDF" to process and display the PDF content.

2. Ask Questions:

   • Enter your query in the textbox labeled "Ask me anything about the content of the PDF".
   • Click "Submit Query" to get answers based on the PDF content.

Code Breakdown

Main Functions

• displayPDF(file): Converts the uploaded PDF into a base64-encoded string for displaying within an iframe.

• data_ingestion(): Loads documents from the data directory and creates a vector store index for persistent storage.

• handle_query(query, lang): Loads the stored index, creates a query engine, and processes the query to generate an answer.

• process_file(uploaded_file): Handles file upload, saves the file to the data directory, and initiates data ingestion.

• chat(file, query, history): Manages the chat interface, processes the uploaded file, detects query language, and appends responses to the chat history.

Gradio Interface

The Gradio interface is built using the gr.Blocks module, with separate columns for file upload and text input. The responses are displayed in a chatbox.

```python with gr.Blocks(css=".gradio-container {background-color: #f0f0f0;} .input-container {margin-bottom: 20px;}") as demo: gr.Markdown("# (PDF) Information and Inference🗞️") gr.Markdown("Retrieval-Augmented Generation") gr.Markdown("Start chat ...🚀")

with gr.Row(equal_height=True):
    with gr.Column():
        file_input = gr.File(label="Upload your PDF Files", type="binary")
        file_submit_button = gr.Button("Upload PDF", variant="primary")

    with gr.Column():
        text_input = gr.Textbox(label="Ask me anything about the content of the PDF:")
        submit_button = gr.Button("Submit Query", variant="secondary")

pdf_output = gr.HTML(label="PDF Display")
chatbox = gr.Chatbot(label="Chat History")

file_submit_button.click(process_file, inputs=[file_input], outputs=[pdf_output])
submit_button.click(chat, inputs=[file_input, text_input, chatbox], outputs=[pdf_output, chatbox])

```

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributions

Feel free to fork the project, make improvements, and submit pull requests. Your contributions are welcome!

Support

For any issues or questions, please open an issue on the GitHub repository or contact the maintainer.

Happy coding! 🚀