app.py

""" Simple Chatbot
@author: Nigel Gebodh
@email: [email protected]

"""
import numpy as np
import streamlit as st
from openai import OpenAI
import os
from dotenv import load_dotenv

load_dotenv()

# Initialize the client
client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1",
    api_key=os.environ.get('HUGGINGFACEHUB_API_TOKEN')  # Replace with your token
)

# Function to reset conversation
def reset_conversation():
    st.session_state.conversation = []
    st.session_state.messages = []
    return None

# Define classification options
classification_types = ["Sentiment Analysis", "Binary Classification", "Multi-Class Classification"]

# Start with a selection between data generation or labeling
st.sidebar.write("Choose Task:")
task = st.sidebar.radio("Do you want to generate data or label data?", ("Data Generation", "Data Labeling"))

# If the user selects Data Labeling
if task == "Data Labeling":
    st.sidebar.write("Choose Classification Type:")
    classification_type = st.sidebar.radio("Select a classification type:", classification_types)

    # Handle Sentiment Analysis
    if classification_type == "Sentiment Analysis":
        st.sidebar.write("Classes: Positive, Negative, Neutral (fixed)")
        class_labels = ["Positive", "Negative", "Neutral"]

    # Handle Binary Classification
    elif classification_type == "Binary Classification":
        class_1 = st.sidebar.text_input("Enter Class 1:")
        class_2 = st.sidebar.text_input("Enter Class 2:")
        class_labels = [class_1, class_2]

    # Handle Multi-Class Classification
    elif classification_type == "Multi-Class Classification":
        class_labels = []
        for i in range(1, 11):  # Allow up to 10 classes
            label = st.sidebar.text_input(f"Enter Class {i} (leave blank to stop):")
            if label:
                class_labels.append(label)
            else:
                break

    # Domain selection
    st.sidebar.write("Specify the Domain:")
    domain = st.sidebar.radio("Choose a domain:", ("Restaurant Reviews", "E-commerce Reviews", "Custom"))
    if domain == "Custom":
        domain = st.sidebar.text_input("Enter Custom Domain:")

    # Specify example length
    st.sidebar.write("Specify the Length of Examples:")
    min_words = st.sidebar.number_input("Minimum word count (10 to 90):", 10, 90, 10)
    max_words = st.sidebar.number_input("Maximum word count (10 to 90):", min_words, 90, 50)

    # Few-shot examples option
    use_few_shot = st.sidebar.radio("Do you want to use few-shot examples?", ("Yes", "No"))
    few_shot_examples = []
    if use_few_shot == "Yes":
        num_examples = st.sidebar.number_input("How many few-shot examples? (1 to 5)", 1, 5, 1)
        for i in range(num_examples):
            example_text = st.text_area(f"Enter example {i+1}:")
            example_label = st.selectbox(f"Select the label for example {i+1}:", class_labels)
            few_shot_examples.append({"text": example_text, "label": example_label})

    # Generate the system prompt based on classification type
    if classification_type == "Sentiment Analysis":
        system_prompt = f"You are a propositional sentiment analysis expert. Your role is to generate sentiment analysis reviews based on the data entered and few-shot examples provided, if any, for the domain '{domain}'."
    elif classification_type == "Binary Classification":
        system_prompt = f"You are an expert in binary classification. Your task is to label examples for the domain '{domain}' with either '{class_1}' or '{class_2}', based on the data provided."
    else:  # Multi-Class Classification
        system_prompt = f"You are an expert in multi-class classification. Your role is to label examples for the domain '{domain}' using the provided class labels."

    st.sidebar.write("System Prompt:")
    st.sidebar.write(system_prompt)

    # Step-by-step thinking
    st.sidebar.write("Generated Data:")
    st.sidebar.write("Think step by step to ensure accuracy in classification.")

    # Accept user input for generating or labeling data
    if prompt := st.chat_input(f"Hi, I'm ready to help with {classification_type} for {domain}. Ask me a question or provide data to classify."):

        # Display user message in chat message container
        with st.chat_message("user"):
            st.markdown(prompt)
        # Add user message to chat history
        st.session_state.messages.append({"role": "user", "content": prompt})

        # Display assistant response in chat message container
        with st.chat_message("assistant"):

            try:
                # Stream the response from the model
                stream = client.chat.completions.create(
                    model="meta-llama/Meta-Llama-3-8B-Instruct",
                    messages=[
                        {"role": m["role"], "content": m["content"]}
                        for m in st.session_state.messages
                    ],
                    temperature=0.5,
                    stream=True,
                    max_tokens=3000,
                )

                response = st.write_stream(stream)

            except Exception as e:
                response = "😵‍💫 Something went wrong. Try again later."
                st.write(response)

        st.session_state.messages.append({"role": "assistant", "content": response})

# If the user selects Data Generation
else:
    st.sidebar.write("This feature will allow you to generate new data. Coming soon!")