app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

# Select the models you want to offer for chat
MODELS = [
    "gpt2",
    "distilgpt2",
    "openai-gpt",
    "openai-gpt-2",
    "openai-gpt3",
]

# Define the system prompt
SYSTEM_PROMPT = "You are a helpful assistant. Answer the user's questions as best as you can."

# Create a dictionary to store conversation history
conversation_history = {}

# Create a function to generate the chatbot response
def chatbot_response(input_text, model_name, system_prompt):
    # Load the selected model and tokenizer
    model = AutoModelForCausalLM.from_pretrained(model_name)
    tokenizer = AutoTokenizer.from_pretrained(model_name)

    # Define the chatbot pipeline
    chatbot_pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

    # Prepare the input for the chatbot
    inputs = tokenizer([SYSTEM_PROMPT + " " + input_text], return_tensors="pt")

    # Generate a response from the chatbot
    response = chatbot_pipe(inputs)

    # Return the response
    return response[0]['generated_text'].strip()

# Create a Gradio interface
interface = gr.Interface(
    fn=chatbot_response,
    inputs=[
        gr.inputs.Textbox(label="User input"),
        gr.inputs.Radio(choices=MODELS, label="Model"),
        gr.inputs.Textbox(label="System prompt", value=SYSTEM_PROMPT),
    ],
    outputs="text",
    title="Large Language Model Chatbot",
    description="Chat with a large language model from the HuggingFace Transformers library.",
)

# Initialize the conversation history
for model in MODELS:
    conversation_history[model] = []

# Define a function to update the conversation history
def update_history(history, new_message):
    history.append(new_message)
    return history

# Define a function to display the conversation history
def display_history(history):
    return "\n".join(history)

# Create a Gradio block to display the conversation history
history_block = gr.Block(
    label="Conversation History",
    elem_id="history",
    visible=False,
)

# Update the conversation history when a new message is sent
def update_history_on_message(history, model, new_message):
    history = update_history(history, new_message)
    conversation_history[model] = history
    return history

# Display the conversation history
def display_history_on_message(history):
    return display_history(history)

# Define event handlers for the Gradio interface
interface.change(history_block.update, [conversation_history], queue=False)
interface.submit(update_history_on_message, [conversation_history], [conversation_history], queue=False)
history_block.change(display_history_on_message, [conversation_history], queue=False)

# Launch the Gradio interface
interface.launch()