Update app.py

app.py

@@ -34,102 +34,61 @@
 
 
 import gradio as gr
-from tools import run_agentic_rag  # 从tool模块导入工作流运行函数
-import time
-
-def format_history(history):
-    """格式化执行历史用于显示"""
-    formatted = []
-    for entry in history:
-        step = entry.get("step", "")
-        status = entry.get("status", "")
-        action = entry.get("action", "")
-        details = ""
-        
-        if step == "检索" and status == "完成":
-            docs = entry.get("documents", [])
-            details = f"检索到 {len(docs)} 个文档"
-        
-        elif step == "生成" and status == "完成":
-            answer = entry.get("answer", "")
-            details = f"生成答案: {answer[:100]}..." if len(answer) > 100 else f"生成答案: {answer}"
-        
-        elif step == "验证" and status == "完成":
-            verif = entry.get("verification", {})
-            valid = verif.get("valid", False)
-            feedback = verif.get("feedback", "")
-            details = f"结果: {'通过' if valid else '失败'}, 反馈: {feedback}"
-        
-        elif step == "准备重试" and status == "完成":
-            feedback = entry.get("feedback", "")
-            details = f"反馈: {feedback}"
-        
-        elif "action" in entry:
-            details = entry["action"]
-        
-        formatted.append(f"{step}: {status} {details}")
-    
-    return "\n".join(formatted)
+import random
+from langgraph.graph import StateGraph, END
+from langchain_core.messages import BaseMessage, HumanMessage, AIMessage, ToolMessage
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_openai import ChatOpenAI
 
-def process_query(question):
-    """处理用户查询"""
-    start_time = time.time()
-    
-    # 执行RAG工作流
-    result = run_agentic_rag(question)
-    
-    # 准备结果显示
-    documents = "\n\n".join([
-        f"文档 {i+1}:\n{doc[:200]}..." if len(doc) > 200 else f"文档 {i+1}:\n{doc}"
-        for i, doc in enumerate(result["documents"])
-    ])
-    
-    end_time = time.time()
-    process_time = f"{end_time - start_time:.2f}秒"
-    
-    return {
-        "answer": result["answer"],
-        "documents": documents,
-        "history": format_history(result["history"]),
-        "stats": f"重试次数: {result['retries_used']} | 处理时间: {process_time}"
-    }
-
-# 创建Gradio界面
-with gr.Blocks(title="Agentic RAG with LangGraph") as demo:
-    gr.Markdown("# 🧠 Agentic RAG 系统 (LangGraph 实现)")
-    gr.Markdown("使用LangGraph实现的带有自我验证和重试机制的RAG系统")
+# Import our custom tools from their modules
+from tools import DuckDuckGoSearchTool, WeatherInfoTool, HubStatsTool
+from retriever import load_guest_dataset
+
+# 定义状态对象
+class AgentState(TypedDict):
+    messages: List[BaseMessage]  # 消息历史
+    plan: Optional[str]  # 当前计划
+    tool_results: List[dict]  # 工具执行结果
+    step_count: int  # 当前步骤计数
+
+# Initialize the model - 使用ChatOpenAI替代HfApiModel
+model = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.7)
+
+# Initialize the tools
+search_tool = DuckDuckGoSearchTool()
+weather_info_tool = WeatherInfoTool()
+hub_stats_tool = HubStatsTool()
+guest_info_tool = load_guest_dataset()
+
+# 所有可用工具
+TOOLS = {
+    "duckduckgo_search": search_tool,
+    "weather_info": weather_info_tool,
+    "hub_stats": hub_stats_tool,
+    "guest_info": guest_info_tool,
+    # 可以在这里添加基础工具
+}
+
+# ========================
+# LangGraph 节点函数
+# ========================
+
+def plan_node(state: AgentState):
+    """规划节点 - 决定下一步行动"""
+    messages = state["messages"]
+    step_count = state["step_count"]
     
-    with gr.Row():
-        with gr.Column():
-            question = gr.Textbox(
-                label="输入问题",
-                placeholder="在此输入您的问题...",
-                lines=3
-            )
-            submit_btn = gr.Button("提交", variant="primary")
+    # 每3步进行规划
+    if step_count % 3 == 0:
+        # 创建规划提示
+        prompt = ChatPromptTemplate.from_messages([
+            ("system", "你是一个智能助手Alfred。根据对话历史和当前状态，规划下一步行动。"),
+            ("human", "当前对话历史:\n{messages}\n\n请规划下一步行动。")
+        ])
         
-        with gr.Column():
-            answer = gr.Textbox(label="最终答案", interactive=False, lines=5)
-            documents = gr.Textbox(label="相关文档", interactive=False, lines=10)
-            history = gr.Textbox(label="执行历史", interactive=False, lines=15)
-            stats = gr.Textbox(label="统计信息", interactive=False)
-    
-    # 事件处理
-    submit_btn.click(
-        fn=process_query,
-        inputs=[question],
-        outputs=[answer, documents, history, stats]
-    )
-    
-    # 示例问题
-    gr.Examples(
-        examples=[
-            "量子纠缠是什么？它有哪些实际应用？",
-            "解释Transformer架构的核心创新点",
-            "如何在PyTorch中实现一个简单的神经网络？"
-        ],
-        inputs=question
-    )
-
-if __name__ == "__main__":
-    demo.launch()
+        chain = prompt | model
+        response = chain.invoke({"messages": "\n".join([m.content for m in messages])})
+        
+        return {
+            "plan": response.content,
+            "messages": messages