app.py

import gradio as gr
from openai import OpenAI
import os

# Retrieve the access token from the environment variable
ACCESS_TOKEN = os.getenv("HF_TOKEN")
print("Access token loaded.")

# Initialize the OpenAI client with the Hugging Face Inference API endpoint
client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1/",
    api_key=ACCESS_TOKEN,
)
print("OpenAI client initialized.")

# We'll define a list of placeholder featured models for demonstration.
# In real usage, replace them with actual model names available on Hugging Face.
models_list = [
    "meta-llama/Llama-3.1-8B-Instruct",
    "microsoft/Phi-3.5-mini-instruct",
    "mistralai/Mistral-7B-Instruct-v0.3",
    "Qwen/Qwen2.5-72B-Instruct"
]

def filter_featured_models(search_term):
    """
    Filters the 'models_list' based on text entered in the search box.
    Returns a gr.update object that changes the choices available
    in the 'featured_models_radio'.
    """
    filtered = [m for m in models_list if search_term.lower() in m.lower()]
    return gr.update(choices=filtered)


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
    frequency_penalty,
    seed,
    custom_model,
    selected_model
):
    """
    This function handles the chatbot response. It takes in:
    - message: the user's new message
    - history: the list of previous messages, each as a tuple (user_msg, assistant_msg)
    - system_message: the system prompt
    - max_tokens: the maximum number of tokens to generate in the response
    - temperature: sampling temperature
    - top_p: top-p (nucleus) sampling
    - frequency_penalty: penalize repeated tokens in the output
    - seed: a fixed seed for reproducibility; -1 will mean 'random'
    - custom_model: a custom Hugging Face model name (if any)
    - selected_model: a model name chosen from the featured models radio button
    """

    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"Frequency Penalty: {frequency_penalty}, Seed: {seed}")
    print(f"Custom model: {custom_model}")
    print(f"Selected featured model: {selected_model}")

    # Convert seed to None if -1 (meaning random)
    if seed == -1:
        seed = None

    # Construct the messages array required by the API
    messages = [{"role": "system", "content": system_message}]

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

    # Append the latest user message
    messages.append({"role": "user", "content": message})

    # Decide which model to use:
    # 1) If the user provided a custom model, use it.
    # 2) Else if they chose a featured model, use it.
    # 3) Otherwise, fall back to a default model.
    if custom_model.strip() != "":
        model_to_use = custom_model.strip()
    elif selected_model is not None and selected_model.strip() != "":
        model_to_use = selected_model.strip()
    else:
        model_to_use = "meta-llama/Llama-3.3-70B-Instruct"  # Default fallback

    print(f"Model selected for inference: {model_to_use}")

    # Start with an empty string to build the response as tokens stream in
    response = ""
    print("Sending request to OpenAI API.")

    # Make the streaming request to the HF Inference API via openai-like client
    for message_chunk in client.chat.completions.create(
        model=model_to_use,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
        frequency_penalty=frequency_penalty,
        seed=seed,
        messages=messages,
    ):
        # Extract the token text from the response chunk
        token_text = message_chunk.choices[0].delta.content
        print(f"Received token: {token_text}")
        response += token_text
        # Yield the partial response to Gradio so it can display in real-time
        yield response

    print("Completed response generation.")


########################
# GRADIO APP LAYOUT
########################

# We’ll build a custom Blocks layout so we can have:
#  - A Featured Models accordion with a search box
#  - Our ChatInterface to handle the conversation
#  - Additional sliders and textboxes for settings (like the original code)
########################

with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
    gr.Markdown("## Serverless Text Generation Hub")
    gr.Markdown(
        "An all-in-one UI for chatting with text-generation models on Hugging Face's Inference API."
    )

    # We keep a Chatbot component for the conversation display
    chatbot = gr.Chatbot(height=600, label="Chat Preview")

    # Textbox for system message
    system_message_box = gr.Textbox(
        value="",
        label="System Message",
        placeholder="Enter a system prompt if you want (optional).",
    )

    # Slider for max_tokens
    max_tokens_slider = gr.Slider(
        minimum=1,
        maximum=4096,
        value=512,
        step=1,
        label="Max new tokens",
    )

    # Slider for temperature
    temperature_slider = gr.Slider(
        minimum=0.1,
        maximum=4.0,
        value=0.7,
        step=0.1,
        label="Temperature",
    )

    # Slider for top_p
    top_p_slider = gr.Slider(
        minimum=0.1,
        maximum=1.0,
        value=0.95,
        step=0.05,
        label="Top-P",
    )

    # Slider for frequency penalty
    freq_penalty_slider = gr.Slider(
        minimum=-2.0,
        maximum=2.0,
        value=0.0,
        step=0.1,
        label="Frequency Penalty",
    )

    # Slider for seed
    seed_slider = gr.Slider(
        minimum=-1,
        maximum=65535,  # Arbitrary upper limit for demonstration
        value=-1,
        step=1,
        label="Seed (-1 for random)",
    )

    # Custom Model textbox
    custom_model_box = gr.Textbox(
        value="",
        label="Custom Model",
        info="(Optional) Provide a custom Hugging Face model path. This will override the selected Featured Model if not empty."
    )

    # Accordion for featured models
    with gr.Accordion("Featured Models", open=False):
        # Textbox for filtering the featured models
        model_search_box = gr.Textbox(
            label="Filter Models",
            placeholder="Search for a featured model...",
            lines=1,
        )
        # Radio for selecting the desired model
        featured_models_radio = gr.Radio(
            label="Select a featured model below",
            choices=models_list,  # Start with the entire list
            value=None,           # No default
            interactive=True
        )

        # We connect the model_search_box to the filter function
        model_search_box.change(
            filter_featured_models,
            inputs=model_search_box,
            outputs=featured_models_radio
        )

    # Now we create our ChatInterface
    # We pass all the extra components as additional_inputs
    interface = gr.ChatInterface(
        fn=respond,
        chatbot=chatbot,
        additional_inputs=[
            system_message_box,
            max_tokens_slider,
            temperature_slider,
            top_p_slider,
            freq_penalty_slider,
            seed_slider,
            custom_model_box,
            featured_models_radio
        ],
        theme="Nymbo/Nymbo_Theme",
        title="Serverless TextGen Hub with Featured Models",
        description=(
            "Use the sliders and textboxes to control generation parameters. "
            "Pick a model from 'Featured Models' or specify a custom model path."
        ),
        # Fill the screen height
        fill_height=True
    )

# If you want the script to be directly executable, launch the demo here:
if __name__ == "__main__":
    print("Launching the demo application...")
    demo.launch()