import gradio as gr
from huggingface_hub import InferenceClient
import base64
from io import BytesIO
from PIL import Image

"""
Hugging Face Hubの推論APIについての詳細は、以下のドキュメントを参照してください: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
client = InferenceClient("Sakalti/SabaVL1-2B")  # モデル名をQwen2-VL-2B-Instructに更新

def encode_image(image):
    buffered = BytesIO()
    image.save(buffered, format="JPEG")
    img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
    return f"data:image/jpeg;base64,{img_str}"

def respond(
    message,
    image,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    if history is None:
        history = []

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

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    if image is not None:
        image_url = encode_image(image)
        messages.append({"role": "user", "content": [{"type": "image_url", "image_url": {"url": image_url}}]})

    messages.append({"role": "user", "content": [{"type": "text", "text": message}]})

    response = ""

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content

        response += token
        yield response

"""
gradioのChatInterfaceのカスタマイズについては、以下のドキュメントを参照してください: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Image(type="pil", label="画像をアップロード"),
        gr.Textbox(value="あなたは親切なチャットボットです。", label="システムメッセージ"),
        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)",
        ),
    ],
)

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