app.py

import gradio as gr
from openai import OpenAI
import os
import json
from datetime import datetime
from zoneinfo import ZoneInfo
import uuid
from pathlib import Path
from huggingface_hub import CommitScheduler


openai_api_key = os.getenv('api_key')
model_name = "gpt-4o-mini"

client = OpenAI(
    api_key=openai_api_key,
)

# Define the file where to save the data. Use UUID to make sure not to overwrite existing data from a previous run.
feedback_file = Path("user_feedback/") / f"data_{uuid.uuid4()}.json"
feedback_folder = feedback_file.parent

# Schedule regular uploads. Remote repo and local folder are created if they don't already exist.
scheduler = CommitScheduler(
    repo_id="misdelivery/demo-test-data",  # Replace with your actual repo ID
    repo_type="dataset",
    folder_path=feedback_folder,
    path_in_repo="data",
    every=1,  # Upload every 1 minutes
)

def save_or_update_conversation(conversation_id, message, response, message_index, liked=None):
    """
    Save or update conversation data in a JSON Lines file.
    If the entry already exists (same id and message_index), update the 'label' field.
    Otherwise, append a new entry.
    """
    with scheduler.lock:
        # Read existing data
        data = []
        if feedback_file.exists():
            with feedback_file.open("r") as f:
                data = [json.loads(line) for line in f if line.strip()]

        # Find if an entry with the same id and message_index exists
        entry_index = next((i for i, entry in enumerate(data) if entry['id'] == conversation_id and entry['message_index'] == message_index), None)

        if entry_index is not None:
            # Update existing entry
            data[entry_index]['label'] = liked
        else:
            # Append new entry
            data.append({
                "id": conversation_id,
                "timestamp": datetime.now(ZoneInfo("Asia/Tokyo")).isoformat(),
                "prompt": message,
                "completion": response,
                "message_index": message_index,
                "label": liked
            })

        # Write updated data back to file
        with feedback_file.open("w") as f:
            for entry in data:
                f.write(json.dumps(entry) + "\n")

def respond(message, history, conversation_id, max_tokens, temperature, top_p):
    messages = [
        {"role": "system", "content": "以下は、タスクを説明する指示です。要求を適切に満たす応答を書きなさい。"}
    ]
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

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

    response = ""
    for chunk in client.chat.completions.create(
        model=model_name,
        messages=messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        if chunk.choices[0].delta.content is not None:
            response += chunk.choices[0].delta.content
        yield response
    
    # Save conversation after the full response is generated
    message_index = len(history)
    save_or_update_conversation(conversation_id, message, response, message_index)

def vote(data: gr.LikeData, history, conversation_id):
    """
    Update user feedback (like/dislike) in the local file.
    """
    message_index = data.index[0]
    liked = data.liked
    save_or_update_conversation(conversation_id, None, None, message_index, liked)

def create_conversation_id():
    return str(uuid.uuid4())

description = """
### gpt400-miniとの会話(期間限定での公開)
- 人工知能開発のため､原則として**このChatBotの入出力データは全て著作権フリー(CC0)で公開する**ため､ご注意ください｡著作物､個人情報､機密情報､誹謗中傷などのデータを入力しないでください｡
- 公開中のデータセット　https://huggingface.co./datasets/misdelivery/demo-test-data　
- **上記の条件に同意する場合のみ**､以下のChatbotを利用してください｡
"""

HEADER = description
FOOTER = """### 注意
- コンテクスト長が4096までなので､あまり会話が長くなると､エラーで停止します｡ページを再読み込みしてください｡
- GPUサーバーが不安定なので､応答しないことがあるかもしれません｡
- この会話データはHugging Face Hubのデータセットに定期的にアップロードされます。"""

def run():
    conversation_id = gr.State(create_conversation_id)
    chatbot = gr.Chatbot(
        elem_id="chatbot",
        scale=1,
        show_copy_button=True,
        height="70%",
        layout="panel",
    )
    with gr.Blocks(fill_height=True) as demo:
        gr.Markdown(HEADER)
        chat_interface = gr.ChatInterface(
            fn=respond,
            stop_btn="Stop Generation",
            cache_examples=False,
            multimodal=False,
            chatbot=chatbot,
            additional_inputs_accordion=gr.Accordion(
                label="Parameters", open=False, render=False
            ),
            additional_inputs=[
                conversation_id,
                gr.Slider(
                    minimum=1,
                    maximum=4096,
                    step=1,
                    value=1024,
                    label="Max tokens",
                    visible=True,
                    render=False,
                ),
                gr.Slider(
                    minimum=0,
                    maximum=1,
                    step=0.1,
                    value=0.3,
                    label="Temperature",
                    visible=True,
                    render=False,
                ),
                gr.Slider(
                    minimum=0,
                    maximum=1,
                    step=0.1,
                    value=1.0,
                    label="Top-p",
                    visible=True,
                    render=False,
                ),
            ],
            analytics_enabled=False,
        )
        chatbot.like(vote, [chatbot, conversation_id], None)
        gr.Markdown(FOOTER)
    demo.queue(max_size=256, api_open=True)
    demo.launch(share=True, quiet=True)

if __name__ == "__main__":
    run()