English | Русский

langchain-gigachat

This is a library integration with GigaChat.

Installation

npm install --save langchain-gigachat

Quickstart

Follow these simple steps to get up and running quickly.

Installation

To install the package use following command:

npm install --save langchain-gigachat

Initialization

To initialize chat model:

import { GigaChat } from "langchain-gigachat"
import { Agent } from 'node:https';

const httpsAgent = new Agent({
    rejectUnauthorized: false,
});

const giga = new GigaChat({
    credentials: 'YOUR_AUTHORIZATION_KEY',
    model: 'GigaChat-Max',
    httpsAgent
})

Usage

Use the GigaChat object to generate responses:

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

const messages = [
    new SystemMessage("Translate following messages to portugese"),
    new HumanMessage("Hello, world!"),
];

const resp = await giga.invoke(messages);

console.log(resp.content);

Use the GigaChat object to create embeddings:

import { GigaChatEmbeddings } from "langchain-gigachat";
import { Agent } from 'node:https';

const httpsAgent = new Agent({
    rejectUnauthorized: false,
});

async function main() {
  const embeddings = new GigaChatEmbeddings({
    credentials: 'YOUR_AUTHORIZATION_KEY',
    httpsAgent
  });

  console.log(await embeddings.embedDocuments(["Словасловаслова"]));
}

main();

Now you can use the GigaChat object with LangChainJS's standard primitives to create LLM-applications.