Skip to main content

Friendli

Friendli 通过可扩展、高效的部署选项提升 AI 应用性能并优化成本节省，专为高需求的 AI 工作负载而设计。

本教程将指导你将 Friendli 与 LangChain 集成。

安装配置​

确保安装了 @langchain/community。

:::提示 请参阅安装集成包的一般说明部分。 :::

npm install @langchain/community @langchain/core

登录 Friendli Suite 创建个人访问令牌，并将其设置为 FRIENDLI_TOKEN 环境变量。
你也可以将团队 ID 设置为 FRIENDLI_TEAM 环境变量。

你可以通过选择要使用的模型来初始化 Friendli 的聊天模型。默认模型是 mixtral-8x7b-instruct-v0-1。你可以在 docs.friendli.ai 查看可用模型。

使用方法​

import { Friendli } from "@langchain/community/llms/friendli";

const model = new Friendli({
  model: "mixtral-8x7b-instruct-v0-1", // Default value
  friendliToken: process.env.FRIENDLI_TOKEN,
  friendliTeam: process.env.FRIENDLI_TEAM,
  maxTokens: 18,
  temperature: 0.75,
  topP: 0.25,
  frequencyPenalty: 0,
  stop: [],
});

const response = await model.invoke(
  "Check the Grammar: She dont like to eat vegetables, but she loves fruits."
);

console.log(response);

/*
Correct: She doesn't like to eat vegetables, but she loves fruits
*/

const stream = await model.stream(
  "Check the Grammar: She dont like to eat vegetables, but she loves fruits."
);

for await (const chunk of stream) {
  console.log(chunk);
}

/*
Cor
rect
:
 She
 doesn
...
she
 loves
 fruits
*/

API Reference:

  • Friendli from @langchain/community/llms/friendli

相关内容​

Was this page helpful?


You can also leave detailed feedback on GitHub.