app.py

import gradio as gr
from openai import AzureOpenAI
import os
from dotenv import load_dotenv
import time

def load_environment():
    """Load environment variables."""
    load_dotenv(override=True)

def initialize_openai_client():
    """Initialize the Azure OpenAI client."""
    return AzureOpenAI(
        azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"),
        api_key=os.getenv("AZURE_OPENAI_API_KEY"),
        api_version="2024-10-01-preview"
    )

def create_assistant(client, vector_store_id):
    """Create an assistant with specified configuration."""
    return client.beta.assistants.create(
        model="gpt-4o",
        instructions="指示がない限り、日本語で回答してください。",
        tools=[{
            "type": "file_search",
            "file_search": {"ranking_options": {"ranker": "default_2024_08_21", "score_threshold": 0}}
        }],
        tool_resources={"file_search": {"vector_store_ids": [vector_store_id]}},
        temperature=0
    )

def create_thread():
    """Create a new thread."""
    return client.beta.threads.create()

def clear_thread(state):
    """セッションをリセットし、チャット履歴をクリアする。"""
    state = initialize_session()  # 新しいスレッドを生成
    return [], ""

def get_annotations(msg):
    annotations = msg.content[0].text.annotations
    file_ids = []
    if annotations:
        for annotation in annotations:
            file_id = annotation.file_citation.file_id
            if file_id in file_ids:
                continue

            print("file_id", file_id)

            cited_file = client.files.retrieve(file_id)

            print("filename", cited_file.filename)

            try:
                content = client.files.content(file_id)
            except Exception as e:
                print(e)
                pass

            file_ids.append(file_id)

    return file_ids

def get_chatbot_response(client, thread_id, assistant_id, message):
    """Get chatbot response for a given message."""
    client.beta.threads.messages.create(
        thread_id=thread_id,
        role="user",
        content=message  # Ensure the content is an object with a `text` key
    )

    run = client.beta.threads.runs.create(
        thread_id=thread_id,
        assistant_id=assistant_id
    )

    while run.status in ["queued", "in_progress", "cancelling"]:
        time.sleep(1)
        run = client.beta.threads.runs.retrieve(
            thread_id=thread_id,
            run_id=run.id
        )

    if run.status == "completed":
        messages = client.beta.threads.messages.list(thread_id=thread_id)

        for msg in messages:
            # file_ids = get_annotations(msg)

            main_text = msg.content[0].text.value
            # main_text += "\n> aaa"
            return main_text

    elif run.status == "requires_action":
        # Handle cases where the assistant requires further action
        pass

    return "Unable to retrieve a response."  # Fallback response

def chatbot_response(history, message):
    """Wrapper function to generate chatbot response."""
    global thread
    # Get response from the API
    assistant_response = get_chatbot_response(client, thread.id, assistant.id, message)

    # Update chat history
    history.append({"role": "user", "content": message})
    history.append({"role": "assistant", "content": assistant_response})

    return history, ""


# Load environment variables
load_environment()
client = initialize_openai_client()
vector_store_id = os.getenv("AZURE_OPENAI_VECTOR_STORE_ID")
assistant = create_assistant(client, vector_store_id)

def respond(message, chat_history, state):
    """チャット履歴と状態を更新する。"""
    thread_id = state["thread_id"]

    bot_message = get_chatbot_response(client, thread_id, assistant.id, message)

    chat_history.append({"role": "user", "content": message})
    chat_history.append({"role": "assistant", "content": bot_message})
    return "", chat_history

def initialize_session():
    """セッションごとに独立したスレッドを初期化する。"""
    thread = create_thread()
    return {"thread_id": thread.id}

with gr.Blocks() as demo:
    gr.Markdown("""
        # Azure OpenAI Assistants API x Gradio x Zenn
        This is a Gradio demo of Retrieval-Augmented Generation (RAG) using the Azure OpenAI Assistants API, applied to [Zenn articles](https://zenn.dev/nakamura196).
    """)

    chatbot = gr.Chatbot(type="messages")
    msg = gr.Textbox(placeholder="ここにメッセージを入力してください...")
    state = gr.State(initialize_session)  # セッションごとの状態を初期化
    clear = gr.Button("Clear")

    msg.submit(respond, [msg, chatbot, state], [msg, chatbot])
    clear.click(clear_thread, inputs=[state], outputs=[chatbot, msg])

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