app.py

import streamlit as st
from sentence_transformers import SentenceTransformer
from langchain.text_splitters import RecursiveCharacterTextSplitter
from langchain.vectorstores import FAISS
from transformers import pipeline
from datasets import load_dataset
import torch

# Set up the Streamlit page configuration
st.set_page_config(page_title="Gen AI Lawyers Guide", layout="centered", page_icon="📄")

# Load summarization pipeline model
@st.cache_resource
def load_summarization_pipeline():
    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")  # Use a summarization model
    return summarizer

summarizer = load_summarization_pipeline()

# Load the CaseHOLD dataset from Hugging Face
@st.cache_data
def load_casehold_dataset():
    dataset = load_dataset("lex_glue", "case_hold", split="train")  # Load CaseHOLD dataset
    texts = [item["context"] for item in dataset]
    return " ".join(texts)

# Split text into manageable chunks
@st.cache_data
def get_text_chunks(text):
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
    chunks = text_splitter.split_text(text)
    return chunks

# Initialize embedding model
@st.cache_resource
def load_embedding_model():
    model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
    return model

embedding_model = load_embedding_model()

# Create a FAISS vector store with embeddings
@st.cache_resource
def load_or_create_vector_store(text_chunks):
    embeddings = [embedding_model.encode(text) for text in text_chunks]
    vector_store = FAISS.from_embeddings(embeddings, text_chunks)  # FAISS setup with embeddings
    return vector_store

# Generate summary based on the retrieved text
def generate_summary_with_huggingface(query, retrieved_text):
    summarization_input = f"{query}\n\nRelated information:\n{retrieved_text}"
    max_input_length = 1024
    summarization_input = summarization_input[:max_input_length]
    summary = summarizer(summarization_input, max_length=500, min_length=50, do_sample=False)
    return summary[0]["summary_text"]

# Generate response for user query
def user_input(user_question, vector_store):
    docs = vector_store.similarity_search(user_question)
    context_text = " ".join([doc.page_content for doc in docs])
    return generate_summary_with_huggingface(user_question, context_text)

# Main function to run the Streamlit app
def main():
    st.title("📄 Gen AI Lawyers Guide with CaseHOLD Dataset")

    # Load CaseHOLD dataset
    st.write("Loading the CaseHOLD dataset from Hugging Face's datasets library...")
    raw_text = load_casehold_dataset()
    text_chunks = get_text_chunks(raw_text)
    vector_store = load_or_create_vector_store(text_chunks)

    # User question input
    user_question = st.text_input("Ask a Question:", placeholder="Type your question here...")

    if st.button("Get Response"):
        if not user_question:
            st.warning("Please enter a question before submitting.")
        else:
            with st.spinner("Generating response..."):
                answer = user_input(user_question, vector_store)
                st.markdown(f"**🤖 AI:** {answer}")

if __name__ == "__main__":
    main()