QA_MiniLM.py

import PyPDF2
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np

# Load models
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')

# Function to process PDF
def process_pdf(pdf_file):
    pdf_reader = PyPDF2.PdfReader(pdf_file)
    document_text = ""
    for page in pdf_reader.pages:
        document_text += page.extract_text()
    sentences = document_text.split('. ')
    embeddings = embedding_model.encode(sentences)
    faiss_index = faiss.IndexFlatL2(embeddings.shape[1])
    faiss_index.add(embeddings)
    return sentences, embeddings, faiss_index

# Function to get relevant context
def get_relevant_context(query, faiss_index, sentences, k=3):
    query_vector = embedding_model.encode([query])
    _, I = faiss_index.search(query_vector, k)
    relevant_sentences = [sentences[i] for i in I[0]]
    return ". ".join(relevant_sentences)

from transformers import pipeline

qa_model = pipeline("question-answering", model="distilbert-base-cased-distilled-squad")

def answer_question(query, faiss_index, sentences):
    if not sentences:
        return "Please upload a document first.", ""
    relevant_context = get_relevant_context(query, faiss_index, sentences)
    answer = qa_model(question=query, context=relevant_context)
    return answer['answer'], relevant_context

import gradio as gr

def process_and_answer(pdf_file, query):
    sentences, embeddings, faiss_index = process_pdf(pdf_file)
    answer, context = answer_question(query, faiss_index, sentences)
    return answer, context

with gr.Blocks() as demo:
    gr.Markdown("# Interactive QA Bot")
    pdf_input = gr.File(label="Upload PDF")
    query_input = gr.Textbox(label="Ask a question about the document")
    answer_output = gr.Textbox(label="Answer")
    context_output = gr.Textbox(label="Relevant Context")
    submit_button = gr.Button("Submit")

    submit_button.click(process_and_answer, inputs=[pdf_input, query_input], outputs=[answer_output, context_output])

demo.launch()