Skip to main content

Layerup 安全

Layerup 安全 集成允许您保护对任何 LangChain 大型语言模型（LLM）、LLM 链或 LLM 代理的调用。该 LLM 对象包裹在任意现有的 LLM 对象外围，从而在您的用户和 LLM 之间建立一个安全层。

虽然 Layerup 安全对象被设计为一个 LLM，但它实际上并不是一个 LLM，它只是包裹在一个 LLM 上，使其能够适配底层 LLM 的相同功能。

准备工作​

首先，您需要从 Layerup 官网 获取一个 Layerup 安全账户。

接下来，通过 仪表板 创建一个项目，并复制您的 API 密钥。我们建议将您的 API 密钥放置在项目的环境变量中。

安装 Layerup 安全 SDK：

npm install @layerup/layerup-security

然后安装 LangChain 社区包：

npm install @langchain/community @langchain/core

现在，您就可以开始使用 Layerup 安全保护您的 LLM 调用了！

import {
  LayerupSecurity,
  LayerupSecurityOptions,
} from "@langchain/community/llms/layerup_security";
import { GuardrailResponse } from "@layerup/layerup-security";
import { OpenAI } from "@langchain/openai";

// Create an instance of your favorite LLM
const openai = new OpenAI({
  modelName: "gpt-3.5-turbo",
  openAIApiKey: process.env.OPENAI_API_KEY,
});

// Configure Layerup Security
const layerupSecurityOptions: LayerupSecurityOptions = {
  // Specify a LLM that Layerup Security will wrap around
  llm: openai,

  // Layerup API key, from the Layerup dashboard
  layerupApiKey: process.env.LAYERUP_API_KEY,

  // Custom base URL, if self hosting
  layerupApiBaseUrl: "https://api.uselayerup.com/v1",

  // List of guardrails to run on prompts before the LLM is invoked
  promptGuardrails: [],

  // List of guardrails to run on responses from the LLM
  responseGuardrails: ["layerup.hallucination"],

  // Whether or not to mask the prompt for PII & sensitive data before it is sent to the LLM
  mask: false,

  // Metadata for abuse tracking, customer tracking, and scope tracking.
  metadata: { customer: "[email protected]" },

  // Handler for guardrail violations on the response guardrails
  handlePromptGuardrailViolation: (violation: GuardrailResponse) => {
    if (violation.offending_guardrail === "layerup.sensitive_data") {
      // Custom logic goes here
    }

    return {
      role: "assistant",
      content: `There was sensitive data! I cannot respond. Here's a dynamic canned response. Current date: ${Date.now()}`,
    };
  },

  // Handler for guardrail violations on the response guardrails
  handleResponseGuardrailViolation: (violation: GuardrailResponse) => ({
    role: "assistant",
    content: `Custom canned response with dynamic data! The violation rule was ${violation.offending_guardrail}.`,
  }),
};

const layerupSecurity = new LayerupSecurity(layerupSecurityOptions);
const response = await layerupSecurity.invoke(
  "Summarize this message: my name is Bob Dylan. My SSN is 123-45-6789."
);

API Reference:

相关内容​

Was this page helpful?


You can also leave detailed feedback on GitHub.