“NaVstreche!” Project — Intelligent Online Meeting Recording and Analysis Service

Challenge

The goal was to create a user-friendly and scalable service that enables a bot to join meetings on popular video conferencing platforms, including Google Meet, Yandex Telemost, Zoom, TrueConf, IVA, and Kontur Meet. The service needed to record meetings, automatically generate videos with subtitles and timestamps, and produce reports with key points derived by a neural network. Users should be able to ask the AI questions about meeting content. Recordings are organized by context (personal, corporate, team) using subdomains, and the architecture must scale to handle a large number of users and meetings.

Solution

We developed a web application with an interactive interface and a robust video conferencing bot. The frontend, built on React, TypeScript, Tailwind, RadixUI, Vite, and i18n, includes a personal account for managing meetings and bots. An interactive video player displays videos with subtitles and timestamps, while meeting reports are generated with a single click. An AI-powered chat allows users to clarify meeting details. Recordings are categorized by subdomains for personal, corporate, and team contexts.

The video conferencing bot is a monolithic Node.js application divided into components. The Broker, utilizing Node.js, TypeScript, Knex (MySQL), Redis, BullMQ, Express, and Zod, validates tasks, manages queues, and can interrupt or block recordings. The Worker, based on Node.js, JavaScript, Playwright, Knex, Redis, BullMQ, and Axios, automatically joins meetings via a browser, records videos using FFMPEG and PulseAudio, saves subtitle data, and monitors participant activity. MySQL and Redis provide reliable data storage, with scalability achieved through Worker replication.

The technology stack includes Node.js, TypeScript, Playwright, Express, and Knex for the backend, MySQL, Redis, and BullMQ for data management, FFMPEG and PulseAudio for video and audio processing, and React, TypeScript, Tailwind, Vite, and i18n for the frontend. AI integration handles report generation and question answering.

Results and Benefits

The service supports recording meetings across more than six video conferencing platforms, automatically generating videos, subtitles with timestamps, and AI-driven reports of key points. Its architecture handles dozens to hundreds of meetings simultaneously with high fault tolerance. Users benefit from clear separation of recordings into personal, corporate, and team contexts. The intuitive video player, AI-generated reports, chat for clarifications, and integration with major video conferencing platforms make the service versatile and adaptable.