Chatbot using GPT 2.0

Overview

This project implements a chatbot using the Hugging Face Transformers library and LangChain. It allows users to interact with a conversational AI that can generate responses based on user inputs and maintain conversational context.

Prerequisites

Before running the project, ensure you have the following installed:

• Python 3.10 or higher
• pip

Installation

To install the required packages, run the following commands:

!pip install huggingface_hub
!pip install transformers
!pip install langchain
!pip install chainlit

All dependencies will be handled automatically.

Getting Started

1. Clone the repository or download the project files. 2.Open a terminal and navigate to the project directory.
2. Start the Chainlit application by running:

!chainlit hello

4. Your app will be available at http://localhost:8000.

Usage

1. Import necessary libraries:

import os
import chainlit as cl
from langchain import HuggingFaceHub, PromptTemplate, LLMChain
from getpass import getpass

2. Set up your Hugging Face API token:

HUGGINGFACEHUB_API_TOKEN = getpass("Enter your Hugging Face API token: ")
os.environ['HUGGINGFACEHUB_API_TOKEN'] = HUGGINGFACEHUB_API_TOKEN

3. Choose a model and create a conversational chain:

model_id = "gpt2-medium"  # or other models
conv_model = HuggingFaceHub(
    huggingfacehub_api_token=os.environ['HUGGINGFACEHUB_API_TOKEN'],
    repo_id=model_id,
    model_kwargs={"temperature": 0.8, "max_new_tokens": 200}
)

4. Define the prompt and run the chatbot:

template = """You are a helpful AI assistant that makes stories by completing the query provided by the user: {query}"""
prompt = PromptTemplate(template=template, input_variables=['query'])
conv_chain = LLMChain(llm=conv_model, prompt=prompt, verbose=True)

response = conv_chain.run("Once upon a time in 1947")
print(response)

Features

• Conversational Memory: The chatbot can remember previous interactions, providing a more engaging user experience.
• Customizable Models: Easily switch between different models from Hugging Face.
• User-Friendly Interface: Built with Chainlit for an intuitive chat interface.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Acknowledgments

• Hugging Face for providing powerful models.
• LangChain for simplifying the interaction with language models.
• Chainlit for the chat interface.