app.py

import gradio as gr
from huggingface_hub import InferenceClient

AVAILABLE_MODELS = [
    "openai/gpt-oss-20b",
    "meta-llama/Llama-3.3-70B-Instruct",
    "meta-llama/Llama-3.1-8B-Instruct",
    "Qwen/Qwen2.5-72B-Instruct",
    "Qwen/Qwen2.5-7B-Instruct",
    "mistralai/Mistral-7B-Instruct-v0.3",
    "mistralai/Mixtral-8x7B-Instruct-v0.1",
    "google/gemma-2-27b-it",
    "google/gemma-2-9b-it",
    "hydffgg/HOS-OSS-270M",
    "Hyggshi-AI/HOS-OSS-200M",
]

def respond(
    message,
    history: list[dict[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
    model_name,
    hf_token: gr.OAuthToken,
):
    client = InferenceClient(token=hf_token.token, model=model_name)

    messages = [{"role": "system", "content": system_message}]
    messages.extend(history)
    messages.append({"role": "user", "content": message})

    response = ""

    try:
        # Thử streaming trước
        for chunk in client.chat_completion(
            messages,
            max_tokens=max_tokens,
            stream=True,
            temperature=temperature,
            top_p=top_p,
        ):
            try:
                choices = chunk.choices
                if choices and choices[0].delta.content:
                    response += choices[0].delta.content
                    yield response
            except (AttributeError, IndexError):
                continue

    except Exception as e:
        # Nếu stream lỗi, fallback sang non-streaming
        try:
            result = client.chat_completion(
                messages,
                max_tokens=max_tokens,
                stream=False,
                temperature=temperature,
                top_p=top_p,
            )
            response = result.choices[0].message.content
            yield response
        except Exception as e2:
            yield f"❌ Lỗi: {str(e2)}\n\nModel `{model_name}` có thể không hỗ trợ chat completion."


chatbot = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
        gr.Dropdown(
            choices=AVAILABLE_MODELS,
            value=AVAILABLE_MODELS[0],
            label="🤖 Model",
            info="Chọn model để chat",
        ),
    ],
)

with gr.Blocks() as demo:
    with gr.Sidebar():
        gr.LoginButton()
        gr.Markdown("### ⚙️ Cài đặt")
        gr.Markdown("Đăng nhập để sử dụng các model HuggingFace.")
    chatbot.render()

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