Requests 工具包
我们可以使用 Requests 工具包 来构建生成 HTTP 请求的代理。
有关所有 API 工具包功能和配置的详细文档，请查看 RequestsToolkit 的 API 参考。
⚠️ 安全说明 ⚠️
让模型自行执行现实世界操作存在固有风险。采取预防措施以减轻这些风险
- 确保与工具相关的权限范围狭窄（例如，数据库操作或 API 请求）；
- 在需要时，使用人机协同工作流程。
设置
安装
此工具包位于 langchain-community 包中
%pip install -qU langchain-community
请注意，如果您希望从单个工具运行中获得自动跟踪，也可以通过取消以下注释来设置您的 LangSmith API 密钥
# os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")
# os.environ["LANGSMITH_TRACING"] = "true"
实例化
首先，我们将演示一个最小的示例。
注意：让模型自行执行现实世界操作存在固有风险。为了使用这些工具，我们必须通过设置 allow_dangerous_request=True 来“选择加入”这些风险。这可能对调用不需要的请求很危险。请确保您的自定义 OpenAPI 规范 (yaml) 是安全的，并且与工具相关的权限范围狭窄。
ALLOW_DANGEROUS_REQUEST = True
我们可以使用 JSONPlaceholder API 作为测试环境。
让我们创建（一部分）它的 API 规范
from typing import Any, Dict, Union
import requests
import yaml
def _get_schema(response_json: Union[dict, list]) -> dict:
if isinstance(response_json, list):
response_json = response_json[0] if response_json else {}
return {key: type(value).__name__ for key, value in response_json.items()}
def _get_api_spec() -> str:
base_url = "https://jsonplaceholder.typicode.com"
endpoints = [
"/posts",
"/comments",
]
common_query_parameters = [
{
"name": "_limit",
"in": "query",
"required": False,
"schema": {"type": "integer", "example": 2},
"description": "Limit the number of results",
}
]
openapi_spec: Dict[str, Any] = {
"openapi": "3.0.0",
"info": {"title": "JSONPlaceholder API", "version": "1.0.0"},
"servers": [{"url": base_url}],
"paths": {},
}
# Iterate over the endpoints to construct the paths
for endpoint in endpoints:
response = requests.get(base_url + endpoint)
if response.status_code == 200:
schema = _get_schema(response.json())
openapi_spec["paths"][endpoint] = {
"get": {
"summary": f"Get {endpoint[1:]}",
"parameters": common_query_parameters,
"responses": {
"200": {
"description": "Successful response",
"content": {
"application/json": {
"schema": {"type": "object", "properties": schema}
}
},
}
},
}
}
return yaml.dump(openapi_spec, sort_keys=False)
api_spec = _get_api_spec()
接下来，我们可以实例化工具包。此 API 不需要授权或其他标头
from langchain_community.agent_toolkits.openapi.toolkit import RequestsToolkit
from langchain_community.utilities.requests import TextRequestsWrapper
toolkit = RequestsToolkit(
requests_wrapper=TextRequestsWrapper(headers={}),
allow_dangerous_requests=ALLOW_DANGEROUS_REQUEST,
)
API 参考：RequestsToolkit | TextRequestsWrapper
工具
查看可用的工具
tools = toolkit.get_tools()
tools
[RequestsGetTool(requests_wrapper=TextRequestsWrapper(headers={}, aiosession=None, auth=None, response_content_type='text', verify=True), allow_dangerous_requests=True),
RequestsPostTool(requests_wrapper=TextRequestsWrapper(headers={}, aiosession=None, auth=None, response_content_type='text', verify=True), allow_dangerous_requests=True),
RequestsPatchTool(requests_wrapper=TextRequestsWrapper(headers={}, aiosession=None, auth=None, response_content_type='text', verify=True), allow_dangerous_requests=True),
RequestsPutTool(requests_wrapper=TextRequestsWrapper(headers={}, aiosession=None, auth=None, response_content_type='text', verify=True), allow_dangerous_requests=True),
RequestsDeleteTool(requests_wrapper=TextRequestsWrapper(headers={}, aiosession=None, auth=None, response_content_type='text', verify=True), allow_dangerous_requests=True)]
在代理中使用
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent
llm = ChatOpenAI(model="gpt-3.5-turbo-0125")
system_message = """
You have access to an API to help answer user queries.
Here is documentation on the API:
{api_spec}
""".format(api_spec=api_spec)
agent_executor = create_react_agent(llm, tools, state_modifier=system_message)
API 参考：ChatOpenAI
example_query = "Fetch the top two posts. What are their titles?"
events = agent_executor.stream(
{"messages": [("user", example_query)]},
stream_mode="values",
)
for event in events:
event["messages"][-1].pretty_print()
================================[1m Human Message [0m=================================
Fetch the top two posts. What are their titles?
==================================[1m Ai Message [0m==================================
Tool Calls:
requests_get (call_RV2SOyzCnV5h2sm4WPgG8fND)
Call ID: call_RV2SOyzCnV5h2sm4WPgG8fND
Args:
url: https://jsonplaceholder.typicode.com/posts?_limit=2
=================================[1m Tool Message [0m=================================
Name: requests_get
[
{
"userId": 1,
"id": 1,
"title": "sunt aut facere repellat provident occaecati excepturi optio reprehenderit",
"body": "quia et suscipit\nsuscipit recusandae consequuntur expedita et cum\nreprehenderit molestiae ut ut quas totam\nnostrum rerum est autem sunt rem eveniet architecto"
},
{
"userId": 1,
"id": 2,
"title": "qui est esse",
"body": "est rerum tempore vitae\nsequi sint nihil reprehenderit dolor beatae ea dolores neque\nfugiat blanditiis voluptate porro vel nihil molestiae ut reiciendis\nqui aperiam non debitis possimus qui neque nisi nulla"
}
]
==================================[1m Ai Message [0m==================================
The titles of the top two posts are:
1. "sunt aut facere repellat provident occaecati excepturi optio reprehenderit"
2. "qui est esse"
API 参考
有关所有 API 工具包功能和配置的详细文档，请查看 RequestsToolkit 的 API 参考。