Stream-режим AI-агента: agent.py

solutions/699cc158d6d3a5544a3ed35b_Stream-режим_AI-агента/agent.py

@@ -1,13 +1,12 @@
 import os
 from typing import Dict, Any
 
-from langgraph.graph import StateGraph, START
-from langgraph.prebuilt.tool_executor import ToolExecutorNode
-from langgraph.checkpoint.memory import MemorySaver
 from langchain_ollama import ChatOllama
 from nomic_embed_text import NomicEmbedText
 from langchain.tools import BaseTool
 from langchain.schema import HumanMessage, AIMessage, SystemMessage
+from langchain.agents import ToolExecutor, AgentExecutor, create_openai_tools_agent
+from langchain.prompts import ChatPromptTemplate
 
 # ---------- Настройки LLM и эмбеддингов ----------
 LLM_MODEL = os.getenv("OLLAMA_MODEL", "llama3")
@@ -30,58 +29,18 @@ class DummyTool(BaseTool):
 
 tool = DummyTool()
 tools = [tool]
+tools_dict = {t.name: t for t in tools}
 
 # ---------- Создание агента ----------
-def create_agent() -> StateGraph:
+prompt_template = ChatPromptTemplate.from_messages(
-    graph = StateGraph()
+    [
+        SystemMessage(content="Ты помощник. Используй инструменты при необходимости."),
+        ("human", "{input}"),
+    ]
+)
 
-    def agent(state: Dict[str, Any]) -> Dict[str, Any]:
+agent = create_openai_tools_agent(llm, tools, prompt=prompt_template)
-        messages = state.get("messages", [])
+executor = AgentExecutor(agent=agent, tools=tools, verbose=False)
-        # Добавляем системное сообщение с инструкцией
-        system_msg = SystemMessage(
-            content="Ты помощник. Используй инструменты при необходимости."
-        )
-        all_messages = [system_msg] + messages
-
-        # Запускаем LLM в режиме stream
-        response = llm.invoke(all_messages, stream=True)
-        return {"messages": [response]}
-
-    graph.add_node("agent", agent)
-
-    def tool_executor(state: Dict[str, Any]) -> Dict[str, Any]:
-        last_msg = state["messages"][-1]
-        if hasattr(last_msg, "tool_calls") and last_msg.tool_calls:
-            # Выполняем первый вызов инструмента
-            call = last_msg.tool_calls[0]
-            name = call["name"]
-            args = eval(call["args"])
-            result = tools_dict[name].invoke(args)
-            # Создаём сообщение с результатом
-            tool_msg = AIMessage(
-                content=result,
-                tool_calls=[call],
-            )
-            return {"messages": [tool_msg]}
-        return {}
-
-    graph.add_node("tool_executor", ToolExecutorNode(tools))
-    graph.set_entry_point("agent")
-    graph.add_edge(START, "agent")
-    graph.add_conditional_edges(
-        "agent",
-        lambda x: "messages" in x and hasattr(x["messages"][-1], "tool_calls"),
-        {
-            True: "tool_executor",
-            False: START,
-        },
-    )
-    graph.set_finish_node("agent")
-
-    return graph.compile(checkpointer=MemorySaver())
-
-tools_dict = {t.name: t for t in tools}
-graph = create_agent()
 
 # ---------- Функции форматирования ----------
 def format_message(message) -> str:
@@ -108,7 +67,7 @@ def run_agent(user_input: str):
     step = 1
     # Инициализируем состояние с пользовательским сообщением
     init_state = {"messages": [HumanMessage(content=user_input)]}
-    stream = graph.stream(init_state, stream_mode=["messages", "updates"])
+    stream = executor.stream(init_state, stream_mode=["messages", "updates"])
 
     for chunk_type, chunk_data in stream:
         if chunk_type == "messages":