Update agent.py
Browse files
agent.py
CHANGED
@@ -1,46 +1,90 @@
|
|
1 |
-
from
|
2 |
-
from .
|
3 |
-
from .
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
-
class
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
-
|
17 |
-
|
18 |
-
self.doc_retriever = DocRetriever()
|
19 |
-
self.web_searcher = WebSearcher()
|
20 |
|
21 |
-
#
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
self.doc_retriever
|
26 |
-
]
|
27 |
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
|
|
|
|
|
|
41 |
|
42 |
def __call__(self, question: str) -> str:
|
43 |
-
print(f"Agent received question: {question[:50]}
|
44 |
-
|
45 |
-
|
46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Annotated, Sequence, TypedDict
|
2 |
+
from langchain_community.llms import HuggingFaceHub
|
3 |
+
from langchain_core.messages import BaseMessage, HumanMessage, AIMessage
|
4 |
+
from langgraph.graph import StateGraph, END
|
5 |
+
from langchain_core.agents import AgentAction, AgentFinish
|
6 |
+
from langchain.agents import create_react_agent
|
7 |
+
from langchain import hub
|
8 |
+
from ai_tools import get_tools # 导入自定义工具集
|
9 |
|
10 |
+
class AgentState(TypedDict):
|
11 |
+
messages: Annotated[Sequence[BaseMessage], operator.add]
|
12 |
+
intermediate_steps: Annotated[list, operator.add]
|
13 |
+
|
14 |
+
def build_graph():
|
15 |
+
# 1. 初始化模型 - 使用HuggingFace免费接口
|
16 |
+
llm = HuggingFaceHub(
|
17 |
+
repo_id="mistralai/Mistral-7B-Instruct-v0.2",
|
18 |
+
model_kwargs={"temperature": 0.1, "max_new_tokens": 500}
|
19 |
+
)
|
20 |
+
|
21 |
+
# 2. 创建ReAct代理
|
22 |
+
prompt = hub.pull("hwchase17/react")
|
23 |
+
tools = get_tools() # 从ai_tools获取工具
|
24 |
+
agent = create_react_agent(llm, tools, prompt)
|
25 |
+
|
26 |
+
# 3. 定义节点行为
|
27 |
+
def agent_node(state: AgentState):
|
28 |
+
input = state["messages"][-1].content
|
29 |
+
result = agent.invoke({
|
30 |
+
"input": input,
|
31 |
+
"intermediate_steps": state["intermediate_steps"]
|
32 |
+
})
|
33 |
+
return {"intermediate_steps": [result]}
|
34 |
+
|
35 |
+
def tool_node(state: AgentState):
|
36 |
+
last_step = state["intermediate_steps"][-1]
|
37 |
+
action = last_step[0] if isinstance(last_step, list) else last_step
|
38 |
|
39 |
+
if not isinstance(action, AgentAction):
|
40 |
+
return {"messages": [AIMessage(content="Invalid action format")]}
|
|
|
|
|
41 |
|
42 |
+
# 执行工具调用
|
43 |
+
tool = next((t for t in tools if t.name == action.tool), None)
|
44 |
+
if not tool:
|
45 |
+
return {"messages": [AIMessage(content=f"Tool {action.tool} not found")]}
|
|
|
|
|
46 |
|
47 |
+
observation = tool.run(action.tool_input)
|
48 |
+
return {"messages": [AIMessage(content=observation)]}
|
49 |
+
|
50 |
+
# 4. 构建状态图
|
51 |
+
workflow = StateGraph(AgentState)
|
52 |
+
workflow.add_node("agent", agent_node)
|
53 |
+
workflow.add_node("tool", tool_node)
|
54 |
+
|
55 |
+
# 5. 定义边和条件
|
56 |
+
def route_action(state: AgentState):
|
57 |
+
last_step = state["intermediate_steps"][-1]
|
58 |
+
action = last_step[0] if isinstance(last_step, list) else last_step
|
59 |
|
60 |
+
if isinstance(action, AgentFinish):
|
61 |
+
return END
|
62 |
+
return "tool"
|
63 |
+
|
64 |
+
workflow.set_entry_point("agent")
|
65 |
+
workflow.add_conditional_edges(
|
66 |
+
"agent",
|
67 |
+
route_action,
|
68 |
+
{"tool": "tool", END: END}
|
69 |
+
)
|
70 |
+
workflow.add_edge("tool", "agent")
|
71 |
|
72 |
+
return workflow.compile()
|
73 |
+
|
74 |
+
class BasicAgent:
|
75 |
+
"""LangGraph智能体封装"""
|
76 |
+
def __init__(self):
|
77 |
+
print("BasicAgent initialized.")
|
78 |
+
self.graph = build_graph()
|
79 |
|
80 |
def __call__(self, question: str) -> str:
|
81 |
+
print(f"Agent received question: {question[:50]}...")
|
82 |
+
messages = [HumanMessage(content=question)]
|
83 |
+
result = self.graph.invoke({
|
84 |
+
"messages": messages,
|
85 |
+
"intermediate_steps": []
|
86 |
+
})
|
87 |
+
|
88 |
+
# 提取最终答案
|
89 |
+
final_message = result["messages"][-1].content
|
90 |
+
return final_message.strip()
|