ChatTencentHunyuan
LangChain.js supports the Tencent Hunyuan family of models.
https://cloud.tencent.com/document/product/1729/104753
Setup
- Sign up for a Tencent Cloud account here.
- Create SecretID & SecretKey here.
- Set SecretID and SecretKey as environment variables named
TENCENT_SECRET_ID
andTENCENT_SECRET_KEY
, respectively.
- npm
- Yarn
- pnpm
npm install @langchain/community
yarn add @langchain/community
pnpm add @langchain/community
If you are using LangChain.js in a browser environment, you'll also need to install the following dependencies:
- npm
- Yarn
- pnpm
npm install crypto-js
yarn add crypto-js
pnpm add crypto-js
And then make sure that you import from the web
as shown below.
Usage
Here's an example:
// in nodejs environment
import { ChatTencentHunyuan } from "@langchain/community/chat_models/tencent_hunyuan";
// in browser environment
// import { ChatTencentHunyuan } from "@langchain/community/chat_models/tencent_hunyuan/web";
import { HumanMessage } from "@langchain/core/messages";
import type { LLMResult } from "@langchain/core/outputs";
const messages = [new HumanMessage("Hello")];
// Default model is hunyuan-pro
const hunyuanPro = new ChatTencentHunyuan({
streaming: false,
temperature: 1,
});
let res = await hunyuanPro.invoke(messages);
console.log(res);
/*
AIMessage {
content: 'Hello! How can I help you today?Is there anything I can do for you?',
name: undefined,
additional_kwargs: {},
response_metadata: {
tokenUsage: { totalTokens: 20, promptTokens: 1, completionTokens: 19 }
},
tool_calls: [],
invalid_tool_calls: []
}
*/
// Use hunyuan-lite
const hunyuanLite = new ChatTencentHunyuan({
model: "hunyuan-lite",
streaming: false,
});
res = await hunyuanLite.invoke(messages);
console.log(res);
/*
AIMessage {
content: '你好!很高兴为你提供服务~有什么我可以帮助你的吗?',
name: undefined,
additional_kwargs: {},
response_metadata: {
tokenUsage: { totalTokens: 14, promptTokens: 1, completionTokens: 13 }
},
tool_calls: [],
invalid_tool_calls: []
}
*/
// Use hunyuan-lite with streaming
const hunyuanLiteStream = new ChatTencentHunyuan({
model: "hunyuan-lite",
streaming: true,
temperature: 1,
});
hunyuanLiteStream.invoke(messages, {
callbacks: [
{
handleLLMEnd(output: LLMResult) {
console.log(output);
/*
{
generations: [
[
[Object], [Object],
[Object], [Object],
[Object], [Object],
[Object], [Object],
[Object]
]
],
llmOutput: {
tokenUsage: { totalTokens: 9, promptTokens: 1, completionTokens: 8 }
}
}
*/
},
handleLLMNewToken(token: string) {
console.log(`token: ${token}`);
/*
token: 你好
token: !
token: 很高兴
token: 能
token: 为您
token: 解答
token: 问题
token: 和建议
token: 方案
token: .
token: 如果您
token: 有其他
token: 需要帮助
token: 的地方
token: ,
token:
token: 随时
token: 告诉我
token: 哦
token: ~
token:
*/
},
},
],
});
API Reference:
- ChatTencentHunyuan from
@langchain/community/chat_models/tencent_hunyuan
- HumanMessage from
@langchain/core/messages
- LLMResult from
@langchain/core/outputs
Related
- Chat model conceptual guide
- Chat model how-to guides