Requests 工具組 (Requests Toolkit)

我們可以使用 Requests 工具包來構建產生 HTTP 請求的代理。

如需所有 API 工具包功能和配置的詳細文件，請前往 RequestsToolkit 的 API 參考。

⚠️ 安全注意事項 ⚠️

讓模型自行決定執行真實世界的動作存在固有風險。請採取預防措施來降低這些風險

• 確保與工具相關聯的權限範圍狹窄 (例如，針對資料庫操作或 API 請求)；
• 在需要時，使用人工參與迴圈工作流程。

設定

安裝

此工具包位於 langchain-community 套件中

%pip install -qU langchain-community

請注意，如果您希望從個別工具的執行中獲得自動追蹤，您也可以透過取消下面的註解來設定您的 LangSmith API 金鑰

# os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")
# os.environ["LANGSMITH_TRACING"] = "true"

實例化

首先，我們將展示一個最小範例。

注意：讓模型自行決定執行真實世界的動作存在固有風險。我們必須透過設定 allow_dangerous_request=True 來「選擇加入」這些風險才能使用這些工具。這對於呼叫不需要的請求可能很危險。請確保您的自訂 OpenAPI 規範 (yaml) 是安全的，並且與工具相關聯的權限範圍狹窄。

ALLOW_DANGEROUS_REQUEST = True

我們可以將 JSONPlaceholder API 作為測試場。

讓我們建立 (一個子集) 它的 API 規範

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()

接下來，我們可以實例化工具包。 我們不需要此 API 的授權或其他標頭

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 參考：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)]

• RequestsGetTool
• RequestsPostTool
• RequestsPatchTool
• RequestsPutTool
• RequestsDeleteTool

在代理中使用

from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent

llm = ChatOpenAI(model="gpt-4o-mini")

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 參考：ChatOpenAI | create_react_agent

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()

================================ Human Message =================================

Fetch the top two posts. What are their titles?
================================== Ai Message ==================================
Tool Calls:
  requests_get (call_RV2SOyzCnV5h2sm4WPgG8fND)
 Call ID: call_RV2SOyzCnV5h2sm4WPgG8fND
  Args:
    url: https://jsonplaceholder.typicode.com/posts?_limit=2
================================= Tool Message =================================
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"
  }
]
================================== Ai Message ==================================

The titles of the top two posts are:
1. "sunt aut facere repellat provident occaecati excepturi optio reprehenderit"
2. "qui est esse"

API 參考

如需所有 API 工具包功能和配置的詳細文件，請前往 RequestsToolkit 的 API 參考。

相關

• 工具 概念指南
• 工具 操作指南