如何在运行时传递回调

先决条件

本指南假设您熟悉以下概念

• 回调

在许多情况下，在运行对象时传递处理程序会更有利。当我们使用 CallbackHandlers 通过 `callbacks` 关键字参数执行运行时，这些回调将由参与执行的所有嵌套对象发出。例如，当处理程序传递到代理时，它将用于与代理及其执行中涉及的所有对象（在本例中为工具和 LLM）相关的所有回调。

这使我们不必将处理程序手动附加到每个单独的嵌套对象。以下是如何使用 LangChain 的内置 ConsoleCallbackHandler 的示例。

import { ConsoleCallbackHandler } from "@langchain/core/tracers/console";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { ChatAnthropic } from "@langchain/anthropic";

const handler = new ConsoleCallbackHandler();

const prompt = ChatPromptTemplate.fromTemplate(`What is 1 + {number}?`);
const model = new ChatAnthropic({
  model: "claude-3-sonnet-20240229",
});

const chain = prompt.pipe(model);

await chain.invoke({ number: "2" }, { callbacks: [handler] });

[chain/start] [1:chain:RunnableSequence] Entering Chain run with input: {
  "number": "2"
}
[chain/start] [1:chain:RunnableSequence > 2:prompt:ChatPromptTemplate] Entering Chain run with input: {
  "number": "2"
}
[chain/end] [1:chain:RunnableSequence > 2:prompt:ChatPromptTemplate] [1ms] Exiting Chain run with output: {
  "lc": 1,
  "type": "constructor",
  "id": [
    "langchain_core",
    "prompt_values",
    "ChatPromptValue"
  ],
  "kwargs": {
    "messages": [
      {
        "lc": 1,
        "type": "constructor",
        "id": [
          "langchain_core",
          "messages",
          "HumanMessage"
        ],
        "kwargs": {
          "content": "What is 1 + 2?",
          "additional_kwargs": {},
          "response_metadata": {}
        }
      }
    ]
  }
}
[llm/start] [1:chain:RunnableSequence > 3:llm:ChatAnthropic] Entering LLM run with input: {
  "messages": [
    [
      {
        "lc": 1,
        "type": "constructor",
        "id": [
          "langchain_core",
          "messages",
          "HumanMessage"
        ],
        "kwargs": {
          "content": "What is 1 + 2?",
          "additional_kwargs": {},
          "response_metadata": {}
        }
      }
    ]
  ]
}
[llm/end] [1:chain:RunnableSequence > 3:llm:ChatAnthropic] [766ms] Exiting LLM run with output: {
  "generations": [
    [
      {
        "text": "1 + 2 = 3",
        "message": {
          "lc": 1,
          "type": "constructor",
          "id": [
            "langchain_core",
            "messages",
            "AIMessage"
          ],
          "kwargs": {
            "content": "1 + 2 = 3",
            "tool_calls": [],
            "invalid_tool_calls": [],
            "additional_kwargs": {
              "id": "msg_01SGGkFVbUbH4fK7JS7agerD",
              "type": "message",
              "role": "assistant",
              "model": "claude-3-sonnet-20240229",
              "stop_sequence": null,
              "usage": {
                "input_tokens": 16,
                "output_tokens": 13
              },
              "stop_reason": "end_turn"
            },
            "response_metadata": {
              "id": "msg_01SGGkFVbUbH4fK7JS7agerD",
              "model": "claude-3-sonnet-20240229",
              "stop_sequence": null,
              "usage": {
                "input_tokens": 16,
                "output_tokens": 13
              },
              "stop_reason": "end_turn"
            }
          }
        }
      }
    ]
  ],
  "llmOutput": {
    "id": "msg_01SGGkFVbUbH4fK7JS7agerD",
    "model": "claude-3-sonnet-20240229",
    "stop_sequence": null,
    "usage": {
      "input_tokens": 16,
      "output_tokens": 13
    },
    "stop_reason": "end_turn"
  }
}
[chain/end] [1:chain:RunnableSequence] [778ms] Exiting Chain run with output: {
  "lc": 1,
  "type": "constructor",
  "id": [
    "langchain_core",
    "messages",
    "AIMessage"
  ],
  "kwargs": {
    "content": "1 + 2 = 3",
    "tool_calls": [],
    "invalid_tool_calls": [],
    "additional_kwargs": {
      "id": "msg_01SGGkFVbUbH4fK7JS7agerD",
      "type": "message",
      "role": "assistant",
      "model": "claude-3-sonnet-20240229",
      "stop_sequence": null,
      "usage": {
        "input_tokens": 16,
        "output_tokens": 13
      },
      "stop_reason": "end_turn"
    },
    "response_metadata": {
      "id": "msg_01SGGkFVbUbH4fK7JS7agerD",
      "model": "claude-3-sonnet-20240229",
      "stop_sequence": null,
      "usage": {
        "input_tokens": 16,
        "output_tokens": 13
      },
      "stop_reason": "end_turn"
    }
  }
}

AIMessage {
  lc_serializable: true,
  lc_kwargs: {
    content: "1 + 2 = 3",
    tool_calls: [],
    invalid_tool_calls: [],
    additional_kwargs: {
      id: "msg_01SGGkFVbUbH4fK7JS7agerD",
      type: "message",
      role: "assistant",
      model: "claude-3-sonnet-20240229",
      stop_sequence: null,
      usage: { input_tokens: 16, output_tokens: 13 },
      stop_reason: "end_turn"
    },
    response_metadata: {}
  },
  lc_namespace: [ "langchain_core", "messages" ],
  content: "1 + 2 = 3",
  name: undefined,
  additional_kwargs: {
    id: "msg_01SGGkFVbUbH4fK7JS7agerD",
    type: "message",
    role: "assistant",
    model: "claude-3-sonnet-20240229",
    stop_sequence: null,
    usage: { input_tokens: 16, output_tokens: 13 },
    stop_reason: "end_turn"
  },
  response_metadata: {
    id: "msg_01SGGkFVbUbH4fK7JS7agerD",
    model: "claude-3-sonnet-20240229",
    stop_sequence: null,
    usage: { input_tokens: 16, output_tokens: 13 },
    stop_reason: "end_turn"
  },
  tool_calls: [],
  invalid_tool_calls: []
}

如果模块已存在关联的回调，则这些回调将与运行时传递的任何回调一起运行。

下一步

您现在已经学习了如何在运行时传递回调。

接下来，查看本节中的其他操作指南，例如如何创建自己的 自定义回调处理程序。