LangGraph 1 - 使用 create_react_agent 创建 Agent

代码

from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent

base_url = 'https://api.siliconflow.cn/v1/'
model = 'Pro/deepseek-ai/DeepSeek-R1'
api_key = 'sk-cyhqy...dngbbdxp'

llm_conf = {
    'base_url': base_url, 
    'model': model, 
    'api_key': api_key
}

# llm = ChatOpenAI(**llm_conf) # 配置方法1
llm = ChatOpenAI(model=model, base_url=base_url, api_key=api_key, temperature=0.1)  

resp = llm.invoke('你是谁？')
print('-- resp : ', resp) 

def get_weather(city: str) -> str:  # (1)!
    """Get weather for a given city."""
    return f"It's always sunny in {city}!"

agent = create_react_agent(
        name='agent1',
        model=llm,
        tools=[get_weather], 
    )


resp2 = agent.invoke(
    {"messages": [{"role": "user", "content": "深圳的天气如何？"}]}
)
print('\n\n')  
print(resp2 )  

返回数据

用户输入后，
1、走到大模型，选择工具、生成参数
2、运行工具，返回结果
3、大模型根据结果，总结，返回给用户

{
	'messages': [
    
    // 1 - 用户提问
    HumanMessage(
      content = '深圳的天气如何？', 
      additional_kwargs = {}, 
      response_metadata = {}, 
      id = '7f0cb6c2-fe5b-4664-b38f-f5b32339d885'
    ), 
    
    // 2 - AI 判断、调用工具、传入参数 
    AIMessage(
      name = 'agent1', 
      id = 'run--c0903412-47a7-4fa1-83f4-83a4fb9c0d5b-0', 
      content = '\n我需要查询深圳的天气情况。根据可用工具，我将使用`get_weather`工具，并传入城市参数"深圳"。以下是工具调用： \n\n', 		
      additional_kwargs = {
        'refusal': None, 
        'tool_calls': [{
          'id': '0197677534ab7eed6dac12c2ef76696e',
          'function': {
            'arguments': '{"city": "\\u6df1\\u5733"}',
            'name': 'get_weather'  
          },
          'type': 'function',
          'index': 0
        }] 
      }, 
      response_metadata = {
        
        'model_name': 'Pro/deepseek-ai/DeepSeek-R1',
        'system_fingerprint': '',
        'id': '01976774e9b9d8f74af09bee5c9b8996',
        'service_tier': None,
        'finish_reason': 'tool_calls',
        'logprobs': None,
        
        'token_usage': {
          'completion_tokens': 122,
          'prompt_tokens': 169,
          'total_tokens': 291,
          'completion_tokens_details': {
            'accepted_prediction_tokens': None,
            'audio_tokens': None,
            'reasoning_tokens': 85,
            'rejected_prediction_tokens': None
          },
          'prompt_tokens_details': None
        },
        
      }, 
      
      tool_calls = [{
        'name': 'get_weather',
        'args': {
          'city': '深圳'
        },
        'id': '0197677534ab7eed6dac12c2ef76696e',
        'type': 'tool_call'
      }], 
      usage_metadata = {
        'input_tokens': 169,
        'output_tokens': 122,
        'total_tokens': 291,
        'input_token_details': {},
        'output_token_details': {
          'reasoning': 85
        }
      }
    ), 
    
    // 3 - 工具返回 
    ToolMessage(
      content = "It's always sunny in 深圳!", 
      name = 'get_weather', 
      id = '87fd53b0-52b3-4814-9e68-7343dbf087ff', 
      tool_call_id = '0197677534ab7eed6dac12c2ef76696e'
    ), 
    
    // 4 - AI 总结回复 
    AIMessage(
      name = 'agent1', 
      id = 'run--827c731c-a9b2-43c5-922d-3ebaf3868b2b-0',  
      content = '\n根据查询结果，深圳的天气总是晴朗的！', 
      additional_kwargs = {
        'refusal': None
      }, 
      usage_metadata = {
        'input_tokens': 258,
        'output_tokens': 87,
        'total_tokens': 345,
        'input_token_details': {},
        'output_token_details': {
          'reasoning': 74
        }
    	},  
      
      response_metadata = {
        'token_usage': {
          'completion_tokens': 87,
          'prompt_tokens': 258,
          'total_tokens': 345,
          
          'completion_tokens_details': {
            'accepted_prediction_tokens': None,
            'audio_tokens': None,
            'reasoning_tokens': 74,
            'rejected_prediction_tokens': None
          },
          'prompt_tokens_details': None
        },
        
        'model_name': 'Pro/deepseek-ai/DeepSeek-R1',
        'system_fingerprint': '',
        'id': '019767753611d940e994d87b10b1ab26',
        'service_tier': None,
        'finish_reason': 'stop',  
        'logprobs': None
      }, 
      
  )
]
}

2025-06-13（五）