Package Exports

langchain-yandexgpt
langchain-yandexgpt/index.mjs

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (langchain-yandexgpt) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

LangChain YandexGPT

LangChain-YandexGPT provides an integration of YandexGPT with LangChain for Node.js, enabling text generation and tool calling workflows. It supports API key and folder ID authentication, model selection such as yandexgpt-lite, temperature control, and structured tool invocation with zod schemas, as demonstrated below.

Installation

npm install langchain-yandexgpt

Example with tool:

import { HumanMessage, SystemMessage } from '@langchain/core/messages';
import { tool } from '@langchain/core/tools';
import { LangChainYandexGPT } from 'langchain-yandexgpt';
import { z } from 'zod';

const llm = new LangChainYandexGPT({
  temperature: 0,
  apiKey: YANDEX_GPT_API_KEY,
  folderID: YANDEX_GPT_CATALOG,
  model: 'yandexgpt-lite',
});

const weatherTool = tool(
  async ({ city }) => {
    return `Погода в городе ${city}: солнечно, температура 22°C, влажность 65%, легкий ветер 5 м/с`;
  },
  {
    name: 'get_weather',
    description: 'Получает актуальную информацию о погоде в указанном городе',
    schema: z.object({
      city: z.string().describe('Название города, например "Москва" или "Санкт-Петербург"'),
    }),
  }
);

const modelWithTools = llm.bindTools([weatherTool]);

const response = await modelWithTools.invoke([
  new SystemMessage(
      'Используй инструмент get_weather чтобы получить информацию о погоде',
  ),
  new HumanMessage('Какая погода в Москве?'),
]);

if (response.tool_calls && response.tool_calls.length > 0) {
  const toolResults = [];
  for (const toolCall of response.tool_calls) {
    if (toolCall.name === 'get_weather') {
      const result = await weatherTool.invoke(toolCall.args);
      console.log(`Результат инструмента: ${toolCall.name} с аргументами: ${toolCall.args}:`, result);
    }
  }
}