app.py

import gradio as gr
from huggingface_hub import InferenceClient
import fitz  # PyMuPDF

client = InferenceClient("opennyaiorg/Aalap-Mistral-7B-v0.1-bf16")

def extract_text_from_pdf(pdf_file):
    document = fitz.open(pdf_file.name)
    text = ""
    for page_num in range(len(document)):
        page = document.load_page(page_num)
        text += page.get_text()
    return text

def summarize_pdf(pdf_file, max_tokens, temperature, top_p):
    text = extract_text_from_pdf(pdf_file)
    response = ""
    messages = [{"role": "user", "content": f"Summarize the following text: {text}"}]

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

def ner_pdf(pdf_file, max_tokens, temperature, top_p):
    text = extract_text_from_pdf(pdf_file)
    response = ""
    messages = [{"role": "user", "content": f"Extract named entities from the following text: {text}"}]

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

def qa_pdf(pdf_file, question, max_tokens, temperature, top_p):
    text = extract_text_from_pdf(pdf_file)
    response = ""
    messages = [{"role": "user", "content": f"Answer the question '{question}' based on the following text: {text}"}]

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

with gr.Blocks() as demo:
    gr.Markdown("# NLP Tasks on PDF Documents")

    with gr.Tab("Summarization"):
        pdf_file = gr.File(label="Upload PDF")
        summarize_button = gr.Button("Summarize")
        summary_output = gr.Textbox(label="Summary")
        summarize_button.click(summarize_pdf, inputs=[pdf_file, gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"), gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"), gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")], outputs=summary_output)
    
    with gr.Tab("Named Entity Recognition (NER)"):
        pdf_file = gr.File(label="Upload PDF")
        ner_button = gr.Button("Extract Entities")
        ner_output = gr.JSON(label="Entities")
        ner_button.click(ner_pdf, inputs=[pdf_file, gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"), gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"), gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")], outputs=ner_output)
    
    with gr.Tab("Question Answering"):
        pdf_file = gr.File(label="Upload PDF")
        question_input = gr.Textbox(label="Enter your question")
        qa_button = gr.Button("Get Answer")
        qa_output = gr.Textbox(label="Answer")
        qa_button.click(qa_pdf, inputs=[pdf_file, question_input, gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"), gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"), gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")], outputs=qa_output)

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