app.py

import os
import fitz  # PyMuPDF for parsing PDF
import streamlit as st
from sentence_transformers import SentenceTransformer, util

# Load a pre-trained SentenceTransformer model
model_name = "paraphrase-MiniLM-L6-v2"  
model = SentenceTransformer(model_name)

# Function to extract text from a PDF file
def extract_text_from_pdf(pdf_path):
    text = ""
    with fitz.open(pdf_path) as pdf_document:
        for page_num in range(pdf_document.page_count):
            page = pdf_document.load_page(page_num)
            page_text = page.get_text()
            text += page_text
            yield page_num + 1, page_text  # Return the page number (1-based) and the extracted text

# Function to perform semantic search
def semantic_search(query, documents, top_k=5):
    query_embedding = model.encode(query, convert_to_tensor=True)

    # Convert the list of documents to embeddings
    document_embeddings = model.encode([text for _, text in documents], convert_to_tensor=True)

    # Compute cosine similarity scores of query with documents
    cosine_scores = util.pytorch_cos_sim(query_embedding, document_embeddings)

    # Sort the results in decreasing order
    results = []
    for idx in range(len(cosine_scores)):
        results.append((documents[idx][0], documents[idx][1], cosine_scores[idx].item()))
    results = sorted(results, key=lambda x: x[2], reverse=True)

    return results[:top_k]

def main():
    st.title("Semantic Search on PDF Documents")

    pdf_file = st.file_uploader("Upload a PDF file:", type=["pdf"])
    query = st.text_input("Enter your query:")

    if st.button("Search"):
        if pdf_file:
            pdf_path = os.path.join(os.getcwd(), pdf_file.name)
            with open(pdf_path, "wb") as f:
                f.write(pdf_file.read())

            # Extract text from the PDF along with page numbers
            pdf_text_with_pages = list(extract_text_from_pdf(pdf_path))

            search_results = semantic_search(query, pdf_text_with_pages)
            os.remove(pdf_path)  # Delete the uploaded file after processing

            st.write(f"Search results for query: '{query}'")
            for i, (page_num, result_text, score) in enumerate(search_results, start=1):
                with st.container():
                    st.write(f"Result {i} - Page {page_num}")
                    st.write(f"Score: {score:.2f}")
                    st.write(result_text)

if __name__ == "__main__":
    main()