app.py

import streamlit as st
import pandas as pd
import json
import os
import plotly.express as px
from transformers import pipeline
from datasets import Dataset
from transformers import RagTokenizer, RagRetriever, RagSequenceForGeneration

# Load the LLaMA-based model with RAG
@st.cache(allow_output_mutation=True)
def load_rag_model():
    retriever = RagRetriever.from_pretrained("facebook/rag-token-base", index_name="custom")
    tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-base")
    model = RagSequenceForGeneration.from_pretrained("facebook/rag-token-base", retriever=retriever)
    return model, tokenizer, retriever

model, tokenizer, retriever = load_rag_model()

# Title of the app
st.title("Interactive Insights Chatbot with LLaMA + RAG")

# Step 1: Upload prompt.json file
prompt_file = st.file_uploader("Upload your prompt.json file", type=["json"])
if prompt_file:
    prompt_data = json.load(prompt_file)
    st.write("Prompt JSON loaded successfully!")

# Step 2: Upload CSV file
csv_file = st.file_uploader("Upload your CSV file", type=["csv"])
if csv_file:
    data = pd.read_csv(csv_file)
    st.write("Data Preview:")
    st.dataframe(data.head())

    # Convert the CSV data to a Hugging Face Dataset for retrieval
    dataset = Dataset.from_pandas(data)

# Step 3: Natural language prompt input
user_prompt = st.text_input("Enter your natural language prompt:")

# Step 4: Process the user prompt and generate insights using LLaMA + RAG
if user_prompt and csv_file:
    st.write(f"Processing your prompt: '{user_prompt}'")

    # Tokenize the prompt for LLaMA + RAG
    inputs = tokenizer(user_prompt, return_tensors="pt")

    # Perform retrieval-augmented generation (RAG) by retrieving data from the dataset and generating the response
    generated = model.generate(input_ids=inputs['input_ids'], num_return_sequences=1, num_beams=2)

    # Decode the output from the LLaMA + RAG model
    output = tokenizer.batch_decode(generated, skip_special_tokens=True)

    st.write(f"Insights generated: {output[0]}")

    # Example: if the prompt asks for a plot (like "show sales over time")
    if "plot sales" in user_prompt.lower():
        # Create a bar chart (you can customize based on the prompt)
        fig = px.bar(data, x='Date', y='Sales', title="Sales Over Time")
        st.plotly_chart(fig)
    else:
        st.write("No recognized visual request in the prompt.")