app.py

#!/usr/bin/env python3
"""
Legal RAG Chatbot for Hugging Face Spaces
直接使用原有的chatbot.py，只添加Gradio界面
"""

import os
import gradio as gr
from chatbot import LegalChatbot

# 配置信息 - 从环境变量获取
MILVUS_DB_PATH = os.environ.get("MILVUS_DB_PATH", "./milvus_legal_codes.db")
COLLECTION_NAME = os.environ.get("COLLECTION_NAME", "legal_codes_collection")
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
OPENAI_BASE_URL = os.environ.get("OPENAI_BASE_URL", "")
MODEL_NAME = os.environ.get("MODEL_NAME", "gpt-4o")

# 初始化聊天机器人
def initialize_chatbot():
    if not OPENAI_API_KEY:
        raise ValueError("❌ OPENAI_API_KEY environment variable is required!")
    
    print(f"🔑 Using API key: {OPENAI_API_KEY[:10]}...")
    print(f"🌐 Base URL: {OPENAI_BASE_URL or 'Default OpenAI'}")
    
    return LegalChatbot(
        milvus_db_path=MILVUS_DB_PATH,
        collection_name=COLLECTION_NAME,
        openai_api_key=OPENAI_API_KEY,
        openai_base_url=OPENAI_BASE_URL if OPENAI_BASE_URL else None,
        model_name=MODEL_NAME
    )

# 全局聊天机器人实例
try:
    chatbot = initialize_chatbot()
    print("✅ Chatbot initialized successfully!")
    chatbot_status = f"✅ **Status**: Connected to database with {chatbot.collection_name} collection"
except Exception as e:
    print(f"❌ Failed to initialize chatbot: {e}")
    chatbot = None
    chatbot_status = f"❌ **Status**: Configuration error - {str(e)}"

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    """
    Gradio ChatInterface响应函数
    使用原有chatbot的流式处理功能
    """
    if chatbot is None:
        yield "❌ Chatbot not initialized. Please check the configuration."
        return
    
    try:
        # 重置聊天机器人的对话历史
        chatbot.reset_conversation()
        
        # 设置系统消息
        if system_message.strip():
            chatbot.conversation_history[0]["content"] = system_message
        
        # 添加历史对话到聊天机器人
        for user_msg, assistant_msg in history:
            if user_msg:
                chatbot.conversation_history.append({"role": "user", "content": user_msg})
            if assistant_msg:
                chatbot.conversation_history.append({"role": "assistant", "content": assistant_msg})
        
        # 使用原有的流式处理功能
        response = ""
        for chunk in chatbot.process_message_stream(message):
            if chunk:
                response += chunk
                yield response
                
    except Exception as e:
        print(f"❌ Error in respond: {e}")
        yield f"抱歉，处理您的消息时出现错误：{str(e)}"

# 准备描述信息
base_description = """🤖 **AI法律助手** - 结合向量数据库搜索和大语言模型的智能法律咨询系统

🔍 **核心功能:**
- 智能查询分析 - 自动判断是否需要搜索法律数据库
- 向量相似度搜索 - 基于Milvus的高效法律文档检索  
- RAG增强生成 - 结合搜索结果提供准确回答
- 实时流式回复 - 支持打字机效果的实时响应

💡 **试试这些问题:**
• "What are the fall protection requirements in Ontario construction?"
• "Tell me about employer duties under Canada Labour Code"  
• "Search for information about workplace safety regulations"
• "What are my rights under the Charter of Rights and Freedoms?"
"""

full_description = base_description + f"\n\n{chatbot_status}"

# 创建Gradio ChatInterface
demo = gr.ChatInterface(
    fn=respond,
    title="⚖️ Legal RAG Assistant",
    description=full_description,
    additional_inputs=[
        gr.Textbox(
            value="You are a helpful legal assistant with expertise in Canadian law. You have access to a legal database and should provide accurate, well-sourced legal information. Always cite specific legal sources when possible. Remember to include appropriate disclaimers that this is for informational purposes only and not legal advice.",
            label="System Message",
            lines=3,
            max_lines=5
        ),
        gr.Slider(
            minimum=1, 
            maximum=2048, 
            value=1024, 
            step=1, 
            label="Max new tokens"
        ),
        gr.Slider(
            minimum=0.1, 
            maximum=2.0, 
            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)"
        ),
    ],
    theme=gr.themes.Soft(),
    analytics_enabled=False
)

if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=False,
        show_error=True
    )