Skip to content

Zergus/NVSBot

Repository files navigation

NotVerySmartBot

Description

Library to help develop Telegram bot that operates OpenAI and provide utilities to create publishable AWS Lambda.

Table of Contents

Installation

npm i --save-dev nvsbot-beta

Usage

DevServer and Bot.createBot provide convenient way to setup bot environment via providing config. Otherwise, Bot class allows for more advanced customization, eg. providing own state management class, if DynamoDB is not desired tool.

Local development

Create executable .js file which will be run by node. Library provides DevServer class which will start development server. Endpoint /webhook will be created for dev server, which accepts POST requests. Example of such executable file:

import "dotenv/config";
import { DevServer } from "./server/DevServer";
import moment from "moment";
import TelegramBot from "node-telegram-bot-api";

const telegramBot = new TelegramBot(process.env.TELEGRAM_TOKEN || "", {
  polling: true,
});

new DevServer({
  command: "/command",
  openAIKey: process.env.OPENAI_API_KEY || "",
  dynamoDBTableName: process.env.TABLE_NAME || "",
  allowedChats: process.env.ALLOWED_CHAT_ID?.split(",") || [],
  defaultResponse: "Help message",
  model: "gpt-3.5-turbo-16k-0613",
  telegramBot,
  endOfConversationFn: (message): { a: string } | void => {
    if (message.match(/^\{.*\}$/)) {
      return JSON.parse(message) as { a: string };
    }
  },
  systemPromptFunc: (username = `undefined`) => {
    const dayToday = moment().format("dddd");
    const dateToday = new Date().toLocaleDateString("uk-UA");
    return "System prompt that returns JSON in the end of conversation";
  },
}).onMessage(async (message, bot) => {
  const result = await bot.processMessage(message);
  if (result) {
    const parsedResult = JSON.parse(result);
    console.log(parsedResult);
  }
});

Setup webhook for your bot:

https://api.telegram.org/bot<bot-token>/setWebhook?url=https://<your-host>/webkook

Deploying WebHook bot to AWS Lambda

IMPORTANT: Lambda function should send instant response to Telegram service about receiving message. This is achieved by creating proxy lambda which after sending response invokes Lambda function with actual bot functionality.

Example of proxy index.js handler for proxy Lambda:

import { Handler } from "nvsbot-beta";

export const handler = Handler.createProxyLambda({
  mainLambdaName: "Your main lambda name",
});

Example of index.js handler for actual bot Lambda:

import { Bot, Handler } from "nvsbot-beta";
import TelegramBot from "node-telegram-bot-api";

const telegramBot = new TelegramBot(process.env.TELEGRAM_TOKEN || "", {
  webHook: true,
});

export const handler = Handler.createMainLambda(async (message) => {
  const bot = Bot.createBot(/* configuration */);
  const result = await bot.processMessage(message);
  /* process the result of conversation */
});

Contributing

Contribute by posting issues or ideas to GitHub Issues.

License

APACHE LICENSE, VERSION 2.0