app.py

import gradio as gr
from huggingface_hub import InferenceClient

def respond(message, history, token, model, system_message, max_tokens, temperature, top_p):
    """
    Handle chat responses using the Hugging Face Inference API.
    """
    # Handle token and model defaults
    token = token.strip()
    model = model.strip()
    
    # Default model selection logic
    if not token:
        model = "gpt2"  # Default public model that doesn't require token
        try:
            client = InferenceClient(model=model)
        except Exception as e:
            yield f"Error initializing client: {str(e)}"
            return
    else:
        model = model or "meta-llama/Llama-3.1-8B-Instruct"  # Default private model
        try:
            client = InferenceClient(model=model, token=token)
        except Exception as e:
            yield f"Error initializing client: {str(e)}"
            return

    # Build message history
    messages = [{"role": "system", "content": system_message}]
    for user_msg, assistant_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})
    messages.append({"role": "user", "content": message})

    # Generate response
    response = ""
    try:
        for chunk in client.chat_completion(
            messages,
            max_tokens=max_tokens,
            stream=True,
            temperature=temperature,
            top_p=top_p,
        ):
            if chunk.choices and chunk.choices[0].delta.content:
                response += chunk.choices[0].delta.content
                yield response
    except Exception as e:
        yield f"API Error: {str(e)}"

# Input components
token_input = gr.Textbox(
    type="password", 
    label="HF API Token (leave empty for public models)",
    placeholder="hf_XXXXXXXXXXXX"
)
model_input = gr.Dropdown(
    label="Model Name",
    choices=[
        "gpt2",
        "HuggingFaceH4/zephyr-7b-beta",
        "meta-llama/Llama-3.1-8B-Instruct"
    ],
    value="gpt2"
)

# Chat interface
demo = gr.ChatInterface(
    fn=respond,
    title="HF Model Chat Interface",
    description="Enter token for private models or use public models without token",
    additional_inputs=[
        token_input,
        model_input,
        gr.Textbox(value="You are helpful AI.", label="System Message"),
        gr.Slider(1, 2048, value=512, label="Max Tokens"),
        gr.Slider(0.1, 4.0, value=0.7, label="Temperature"),
        gr.Slider(0.1, 1.0, value=0.95, label="Top-p"),
    ],
)

if __name__ == "__main__":
    demo.launch()