Skip to main content

如何将运行时配置附加到一个 Runnable 上

预备知识

本指南假定您已经熟悉以下概念：

有时我们想要使用预定义的配置来调用一个 Runnable，而无需调用方指定这些配置。我们可以使用 Runnable.withConfig() 方法提前设置这些参数。

绑定停止序列​

假设我们有一个简单的提示词 + 模型链：

:::提示 请参阅安装集成包的一般说明部分。 :::

yarn add @langchain/openai @langchain/core
import { StringOutputParser } from "@langchain/core/output_parsers";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { ChatOpenAI } from "@langchain/openai";

const prompt = ChatPromptTemplate.fromMessages([
  [
    "system",
    "Write out the following equation using algebraic symbols then solve it. Use the format\n\nEQUATION:...\nSOLUTION:...\n\n",
  ],
  ["human", "{equation_statement}"],
]);

const model = new ChatOpenAI({ model: "gpt-4o", temperature: 0 });

const runnable = prompt.pipe(model).pipe(new StringOutputParser());

const res = await runnable.invoke({
  equation_statement: "x raised to the third plus seven equals 12",
});

console.log(res);
EQUATION: x^3 + 7 = 12

SOLUTION:
Subtract 7 from both sides:
x^3 = 5

Take the cube root of both sides:
x = ∛5

在某些提示技术中，设置一个或多个stop（停止）词可能会很有用，当模型生成这些词时将停止生成。在直接使用模型时，我们通过传递给invoke的额外options参数来设置stop词。例如，如果我们想将前面的例子修改为方程格式化工具该怎么办？我们可以指示生成过程在遇到单词SOLUTION时停止，这样输出的结果就只会是格式化后的方程。

在这种情况下，我们将模型作为RunnableSequence的一部分使用。我们无需依赖前面的步骤输出包含停止词的配置，而可以在创建RunnableSequence时使用withConfig直接绑定所需的配置：

// stop generating after the equation is written
const equationFormatter = prompt
  .pipe(model.withConfig({ stop: ["SOLUTION"] }))
  .pipe(new StringOutputParser());

// generate only the equation, without needing to set the stop word
const formattedEquation = await equationFormatter.invoke({
  equation_statement: "x raised to the third plus seven equals 12",
});

console.log(formattedEquation);
EQUATION: x^3 + 7 = 12

这使得生成的 Runnable 管道更易于使用。调用者无需了解所使用的具体提示格式，也无需在调用时指定停止词。他们只需运行自己的查询，就能获得预期的结果。

附加 OpenAI 工具​

另一个常见的用例是工具调用。虽然对于支持工具调用的模型，你通常应该使用 bindTools() 方法，但如果你需要更底层的控制，也可以直接绑定特定于提供者的参数：

const tools = [
  {
    type: "function",
    function: {
      name: "get_current_weather",
      description: "Get the current weather in a given location",
      parameters: {
        type: "object",
        properties: {
          location: {
            type: "string",
            description: "The city and state, e.g. San Francisco, CA",
          },
          unit: { type: "string", enum: ["celsius", "fahrenheit"] },
        },
        required: ["location"],
      },
    },
  },
];

const modelWithTools = new ChatOpenAI({ model: "gpt-4o" }).withConfig({
  tools,
});

await modelWithTools.invoke("What's the weather in SF, NYC and LA?");
AIMessage {
  "id": "chatcmpl-BXjkosti03tvSmaxAuYtpRvbEkhRx",
  "content": "",
  "additional_kwargs": {
    "tool_calls": [
      {
        "id": "call_a15PVBt9g3eCHULn7DBRbL9a",
        "type": "function",
        "function": "[Object]"
      },
      {
        "id": "call_bQrHLyJ6fAaNkEPNBtgYfIFb",
        "type": "function",
        "function": "[Object]"
      },
      {
        "id": "call_FxRswXKWou53G0LNQQCvPod4",
        "type": "function",
        "function": "[Object]"
      }
    ]
  },
  "response_metadata": {
    "tokenUsage": {
      "promptTokens": 82,
      "completionTokens": 71,
      "totalTokens": 153
    },
    "finish_reason": "tool_calls",
    "model_name": "gpt-4o-2024-08-06",
    "usage": {
      "prompt_tokens": 82,
      "completion_tokens": 71,
      "total_tokens": 153,
      "prompt_tokens_details": {
        "cached_tokens": 0,
        "audio_tokens": 0
      },
      "completion_tokens_details": {
        "reasoning_tokens": 0,
        "audio_tokens": 0,
        "accepted_prediction_tokens": 0,
        "rejected_prediction_tokens": 0
      }
    },
    "system_fingerprint": "fp_90122d973c"
  },
  "tool_calls": [
    {
      "name": "get_current_weather",
      "args": {
        "location": "San Francisco, CA"
      },
      "type": "tool_call",
      "id": "call_a15PVBt9g3eCHULn7DBRbL9a"
    },
    {
      "name": "get_current_weather",
      "args": {
        "location": "New York, NY"
      },
      "type": "tool_call",
      "id": "call_bQrHLyJ6fAaNkEPNBtgYfIFb"
    },
    {
      "name": "get_current_weather",
      "args": {
        "location": "Los Angeles, CA"
      },
      "type": "tool_call",
      "id": "call_FxRswXKWou53G0LNQQCvPod4"
    }
  ],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "output_tokens": 71,
    "input_tokens": 82,
    "total_tokens": 153,
    "input_token_details": {
      "audio": 0,
      "cache_read": 0
    },
    "output_token_details": {
      "audio": 0,
      "reasoning": 0
    }
  }
}

下一步​

现在您已经了解了如何将运行时参数绑定到一个 Runnable。

接下来，您可能对我们的如何通过链传递数据的指南感兴趣：通过链传递数据。

Was this page helpful?


You can also leave detailed feedback on GitHub.