Speech recognition via wit.ai (https://wit.ai/docs/http/20210928/#post__speech_link), vk chatbot api (https://vk.com/dev/bots_docs)

wit_ai_vk_group_bot.js

// node ^16.4.2

/**
 * npm i node-vk-sdk@1.1.7 node-fetch@2 --save
 * GROUP_ACCESS_TOKEN=1234 WIT_AI_CLIENT_TOKEN=4567 GROUP_ID=1234 node wit_ai_vk_group_bot.js
 */

const {
  VKApi,
  ConsoleLogger,
  BotsLongPollUpdatesProvider,
} = require("node-vk-sdk");
const fetch = require("node-fetch");

const WIT_AI_SPEECH_API_URL = "https://api.wit.ai/speech?v=20220126";
const NEW_MESSAGE_TYPE = "message_new";
const AUDIO_MESSAGE_TYPE = "audio_message";

const api = new VKApi({
  // https://vk.com/{GROUP_ID}?act=tokens -> Create access token
  // https://vk.com/{GROUP_ID}?act=longpoll_api -> API version 5.126 (for node-vk-sdk 1.1.7)
  // optional:
  // https://vk.com/{GROUP_ID}?act=messages&tab=bots -> "Bot abilities": Enabled, "Ability to add this community to chats" - checked
  token: process.env.GROUP_ACCESS_TOKEN,
  logger: new ConsoleLogger(),
});

const updatesProvider = new BotsLongPollUpdatesProvider(
  api,
  process.env.GROUP_ID
);

updatesProvider.getUpdates((updates) => {
  if (!updates?.length) {
    return;
  }

  updates.forEach((update) => {
    if (update?.type !== NEW_MESSAGE_TYPE) {
      return;
    }

    if (!update?.object?.message?.attachments?.length) {
      return;
    }

    update.object.message.attachments.forEach(async (attachment) => {
      if (attachment?.type !== AUDIO_MESSAGE_TYPE) {
        return;
      }

      const link = attachment?.audio_message?.link_mp3;
      if (!link) {
        return;
      }

      let blob;
      try {
        blob = await fetch(link).then((response) => response.blob());
      } catch (error) {
        api.logger.error("", error);
      }

      if (!blob) {
        return;
      }

      const response = await fetch(WIT_AI_SPEECH_API_URL, {
        method: "POST",
        headers: {
          // https://wit.ai/apps -> "New app" -> "Client access token"
          Authorization: `Bearer ${process.env.WIT_AI_CLIENT_TOKEN}`,
          "Content-Type": "audio/mpeg3",
        },
        body: blob,
      })
        .then((response) => response.body)
        .then(
          (res) =>
            new Promise((resolve) => {
              const chunks = [];

              res.on("readable", () => {
                let chunk;
                while (null !== (chunk = res.read())) {
                  chunks.push(chunk);
                }
              });

              res.on("end", () => {
                resolve(chunks[chunks.length - 1]);
              });
            })
        )
        .catch((error) => {
          api.logger.log("", error);
        });

      try {
        const message = JSON.parse(response.toString())?.text;
        if (!message) {
          return;
        }
        
        await api.messagesSend({
          peer_id: update.object.message.peer_id,
          reply_to: update.object.message.id,
          message,
          random_id: Math.random() * 9999,
        });
      } catch (error) {
        api.logger.error("", error);
      }
    });
  });
});