app.py

import streamlit as st
from langchain.document_loaders.csv_loader import CSVLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.llms import CTransformers
from langchain.chains import ConversationalRetrievalChain
import pandas as pd
from io import StringIO

# Set up the Streamlit page configuration
st.set_page_config(page_title="Conversational AI with CSV Data", page_icon="🤖", layout="wide")

# Sidebar - User can input the CSV file path and query
st.sidebar.title("Upload CSV and Ask Questions")
uploaded_file = st.sidebar.file_uploader("Upload your CSV file", type="csv")

if uploaded_file is not None:
    # Read the CSV file into a Pandas DataFrame
    string_data = StringIO(uploaded_file.getvalue().decode("utf-8"))
    df = pd.read_csv(string_data)

    # Display the first few rows of the DataFrame
    st.write("Data successfully loaded!")
    st.write(df.head())

    # Convert the DataFrame to the format expected by CSVLoader
    df.to_csv("temp.csv", index=False)

    # Load and process the CSV file using the CSVLoader
    loader = CSVLoader(file_path="temp.csv", encoding="utf-8", csv_args={'delimiter': ','})
    data = loader.load()

    # Split the text into chunks
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=20)
    text_chunks = text_splitter.split_documents(data)

    # Display the number of chunks
    st.write(f"Number of text chunks created: {len(text_chunks)}")

    # Download Sentence Transformers Embedding from Hugging Face
    embeddings = HuggingFaceEmbeddings(model_name='sentence-transformers/all-MiniLM-L6-v2')

    # Convert the text chunks into embeddings and save the embeddings into FAISS knowledge base
    docsearch = FAISS.from_documents(text_chunks, embeddings)
    docsearch.save_local("vectorstore/db_faiss")

    # Load the LLM model
    llm = CTransformers(
        model="models/llama-2-7b-chat.ggmlv3.q4_0.bin",
        model_type="llama",
        max_new_tokens=512,
        temperature=0.1
    )

    # Create the conversational retrieval chain
    qa = ConversationalRetrievalChain.from_llm(llm, retriever=docsearch.as_retriever())

    # Chat history
    chat_history = []

    # Input prompt from the user
    user_input = st.text_input("Ask a question about the data:")

    if st.button("Ask"):
        if user_input:
            result = qa({"question": user_input, "chat_history": chat_history})
            st.write(f"Response: {result['answer']}")
            chat_history.append((user_input, result['answer']))
        else:
            st.write("Please enter a question.")

else:
    st.write("Please upload a CSV file to get started.")