app.py

import os
from langchain_core.prompts import ChatPromptTemplate, HumanMessagePromptTemplate, SystemMessage
from langchain.schema import SystemMessage
import streamlit as st
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from langchain_community.llms.huggingface_pipeline import HuggingFacePipeline
import nltk
import json
import pandas as pd

# Download nltk stopwords
nltk.download('stopwords')

# Function to load the conversation history
def load_conversation_history(file):
    with open(file, 'r') as f:
        return json.load(f)

# Function to save the conversation history
def save_conversation_history(history, file):
    with open(file, 'w') as f:
        json.dump(history, f)

# Initialize conversation history
conversation_history = []
if st.session_state.get('conversation_history'):
    conversation_history = st.session_state.conversation_history

# Title
st.title('Culture AI v.0.1')

# Get the Hugging Face access token from the environment variable
HF_TOKEN = os.getenv("HF_TOKEN")

# Model selection
model_name = st.selectbox('Choose a model:', [
    'meta-llama/Llama-3.2-11B-Vision-Instruct',
    'speakleash/Bielik-11B-v2.3-Instruct',  
    # Add your private model here
])

# Upload dataset
dataset_file = st.file_uploader('Upload your dataset (CSV format)', type='csv')
if dataset_file:
    df = pd.read_csv(dataset_file)

# Initialize tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=True)
model = AutoModelForCausalLM.from_pretrained(model_name)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, return_full_text=True)
llm = HuggingFacePipeline(pipeline=pipe)

# Chat interface
st.write('## Chat')
topic = st.text_input('Enter a topic for the conversation:', 'Machine Learning')
prompt = ChatPromptTemplate.from_messages([
    SystemMessage(content=f"Write a response related to the input topic in one paragraph"),
    HumanMessagePromptTemplate.from_template("```{topic}```"),
])

chain = prompt | llm

if st.button('Generate Response'):
    output = chain.invoke({"topic": topic})
    st.write(output.content)
    conversation_history.append({"user": topic, "assistant": output.content})

# Save conversation history
if st.button('Save Conversation History'):
    save_conversation_history(conversation_history, 'conversation_history.json')
    st.success('Conversation history saved!')

# Display conversation history
st.write('## Conversation History')
st.write(conversation_history)

# Update session state for conversation history
st.session_state.conversation_history = conversation_history