如何过滤消息

filterMessages 函数在 @langchain/core 版本 0.2.8 及更高版本中可用。

在更复杂的链和代理中，我们可能会使用消息列表来跟踪 state。此列表可以开始累积来自多个不同模型、发言者、子链等的 消息，我们可能只想将此完整消息列表的子集传递给链/代理中的每次模型调用。

filterMessages 实用程序使您可以轻松地按类型、id 或名称过滤消息。

基本用法

import {
  HumanMessage,
  SystemMessage,
  AIMessage,
  filterMessages,
} from "@langchain/core/messages";

const messages = [
  new SystemMessage({ content: "you are a good assistant", id: "1" }),
  new HumanMessage({ content: "example input", id: "2", name: "example_user" }),
  new AIMessage({
    content: "example output",
    id: "3",
    name: "example_assistant",
  }),
  new HumanMessage({ content: "real input", id: "4", name: "bob" }),
  new AIMessage({ content: "real output", id: "5", name: "alice" }),
];

filterMessages(messages, { includeTypes: ["human"] });

[
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'example input',
      id: '2',
      name: 'example_user',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'example input',
    name: 'example_user',
    additional_kwargs: {},
    response_metadata: {},
    id: '2'
  },
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real input',
      id: '4',
      name: 'bob',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real input',
    name: 'bob',
    additional_kwargs: {},
    response_metadata: {},
    id: '4'
  }
]

filterMessages(messages, {
  excludeNames: ["example_user", "example_assistant"],
});

[
  SystemMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'you are a good assistant',
      id: '1',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'you are a good assistant',
    name: undefined,
    additional_kwargs: {},
    response_metadata: {},
    id: '1'
  },
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real input',
      id: '4',
      name: 'bob',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real input',
    name: 'bob',
    additional_kwargs: {},
    response_metadata: {},
    id: '4'
  },
  AIMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real output',
      id: '5',
      name: 'alice',
      tool_calls: [],
      invalid_tool_calls: [],
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real output',
    name: 'alice',
    additional_kwargs: {},
    response_metadata: {},
    id: '5',
    tool_calls: [],
    invalid_tool_calls: [],
    usage_metadata: undefined
  }
]

filterMessages(messages, {
  includeTypes: [HumanMessage, AIMessage],
  excludeIds: ["3"],
});

[
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'example input',
      id: '2',
      name: 'example_user',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'example input',
    name: 'example_user',
    additional_kwargs: {},
    response_metadata: {},
    id: '2'
  },
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real input',
      id: '4',
      name: 'bob',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real input',
    name: 'bob',
    additional_kwargs: {},
    response_metadata: {},
    id: '4'
  },
  AIMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real output',
      id: '5',
      name: 'alice',
      tool_calls: [],
      invalid_tool_calls: [],
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real output',
    name: 'alice',
    additional_kwargs: {},
    response_metadata: {},
    id: '5',
    tool_calls: [],
    invalid_tool_calls: [],
    usage_metadata: undefined
  }
]

链接

filterMessages 可用于命令式 (如上所示) 或声明式，使您可以轻松地与链中的其他组件进行组合

import { ChatAnthropic } from "@langchain/anthropic";

const llm = new ChatAnthropic({
  model: "claude-3-sonnet-20240229",
  temperature: 0,
});
// Notice we don't pass in messages. This creates
// a RunnableLambda that takes messages as input
const filter_ = filterMessages({
  excludeNames: ["example_user", "example_assistant"],
  end,
});
const chain = filter_.pipe(llm);
await chain.invoke(messages);

AIMessage {
  lc_serializable: true,
  lc_kwargs: {
    content: [],
    additional_kwargs: {
      id: 'msg_01S2LQc1NLhtPHurW3jNRsCK',
      type: 'message',
      role: 'assistant',
      model: 'claude-3-sonnet-20240229',
      stop_reason: 'end_turn',
      stop_sequence: null,
      usage: [Object]
    },
    tool_calls: [],
    usage_metadata: { input_tokens: 16, output_tokens: 3, total_tokens: 19 },
    invalid_tool_calls: [],
    response_metadata: {}
  },
  lc_namespace: [ 'langchain_core', 'messages' ],
  content: [],
  name: undefined,
  additional_kwargs: {
    id: 'msg_01S2LQc1NLhtPHurW3jNRsCK',
    type: 'message',
    role: 'assistant',
    model: 'claude-3-sonnet-20240229',
    stop_reason: 'end_turn',
    stop_sequence: null,
    usage: { input_tokens: 16, output_tokens: 3 }
  },
  response_metadata: {
    id: 'msg_01S2LQc1NLhtPHurW3jNRsCK',
    model: 'claude-3-sonnet-20240229',
    stop_reason: 'end_turn',
    stop_sequence: null,
    usage: { input_tokens: 16, output_tokens: 3 }
  },
  id: undefined,
  tool_calls: [],
  invalid_tool_calls: [],
  usage_metadata: { input_tokens: 16, output_tokens: 3, total_tokens: 19 }
}

查看LangSmith 跟踪，我们可以看到，在将消息传递给模型之前，会对其进行过滤。

仅查看过滤器_，我们可以看到它是一个 Runnable 对象，可以像所有 Runnable 一样调用

await filter_.invoke(messages);

[
  SystemMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'you are a good assistant',
      id: '1',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'you are a good assistant',
    name: undefined,
    additional_kwargs: {},
    response_metadata: {},
    id: '1'
  },
  HumanMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real input',
      id: '4',
      name: 'bob',
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real input',
    name: 'bob',
    additional_kwargs: {},
    response_metadata: {},
    id: '4'
  },
  AIMessage {
    lc_serializable: true,
    lc_kwargs: {
      content: 'real output',
      id: '5',
      name: 'alice',
      tool_calls: [],
      invalid_tool_calls: [],
      additional_kwargs: {},
      response_metadata: {}
    },
    lc_namespace: [ 'langchain_core', 'messages' ],
    content: 'real output',
    name: 'alice',
    additional_kwargs: {},
    response_metadata: {},
    id: '5',
    tool_calls: [],
    invalid_tool_calls: [],
    usage_metadata: undefined
  }
]

API 参考

有关所有参数的完整说明，请访问API 参考。