app.py

#refer llama recipes for more info https://github.com/huggingface/huggingface-llama-recipes/blob/main/inference-api.ipynb
#huggingface-llama-recipes : https://github.com/huggingface/huggingface-llama-recipes/tree/main

import gradio as gr
from openai import OpenAI
import os

ACCESS_TOKEN = os.getenv("myHFtoken")

print("Access token loaded.")

client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1/",
    api_key=ACCESS_TOKEN,
)

print("Client initialized.")

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
    model_name,  # New parameter for model selection
):
    print(f"Received message: {message}")
    print(f"History: {history}")
    print(f"System message: {system_message}")
    print(f"Max tokens: {max_tokens}, Temperature: {temperature}, Top-P: {top_p}")
    print(f"Selected model: {model_name}")

    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
            print(f"Added user message to context: {val[0]}")
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})
            print(f"Added assistant message to context: {val[1]}")

    messages.append({"role": "user", "content": message})

    response = ""
    print("Sending request to OpenAI API.")
    
    for message in client.chat.completions.create(
        model=model_name,  # Use the selected model
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
        messages=messages,
    ):
        token = message.choices[0].delta.content
        print(f"Received token: {token}")
        response += token
        yield response

    print("Completed response generation.")
        
chatbot = gr.Chatbot(height=600)

print("Chatbot interface created.")

# Define the list of models
models = [
    "PowerInfer/SmallThinker-3B-Preview", #OK
    "Qwen/QwQ-32B-Preview", #OK
    "Qwen/Qwen2.5-Coder-32B-Instruct", #OK
    "meta-llama/Llama-3.2-3B-Instruct", #OK
    #"Qwen/Qwen2.5-32B-Instruct", #fail, too large
    #"microsoft/Phi-3-mini-128k-instruct", #fail
    #"microsoft/Phi-3-medium-128k-instruct", #fail
    #"microsoft/phi-4", #fail, too large to be loaded automatically (29GB > 10GB)
    #"meta-llama/Llama-3.3-70B-Instruct", #fail, need HF Pro subscription
]

# Add a title and move the model dropdown to the top
with gr.Blocks() as demo:
    gr.Markdown("# LLM Test (HF API)")  # Add a title to the top of the UI
    
    # Add the model dropdown above the chatbot
    model_dropdown = gr.Dropdown(choices=models, value=models[0], label="Select Model")
    
    # Use the existing ChatInterface
    gr.ChatInterface(
        respond,
        additional_inputs=[
            gr.Textbox(value="", label="System message"),
            gr.Slider(minimum=1, maximum=4096, value=1024, step=1, label="Max new tokens"),
            gr.Slider(minimum=0.1, maximum=1.0, value=0.3, step=0.1, label="Temperature"),
            gr.Slider(
                minimum=0.1,
                maximum=1.0,
                value=0.95,
                step=0.05,
                label="Top-P",
            ),
            model_dropdown,  # Pass the dropdown as an additional input
        ],
        fill_height=True,
        chatbot=chatbot,
    )

print("Gradio interface initialized.")

if __name__ == "__main__":
    print("Launching the demo application.")
    demo.launch()