Hello, Habr!

In my previous articles (one, two, three) I talked about the origin of the idea and the first steps in developing a Telegram bot for managing servers. What started out as a simple Python script for checking uptime and rebooting a couple of personal VPS, in recent months has acquired “meat” and turned into an adult ecosystem with the Agent-Server pattern, its own WebUI, WAF and PWA.

Today I want to share the experience that I gained in the process of large-scale refactoring (from version 1.13.0 to the current 1.21.x), talk about the rake that I stepped on when working with memory and network, and compare my product with what is already on the market.

I’ll make a reservation right away: although I am listed as the only developer, the project is not created in a vacuum. First, I actively use AI tools to accelerate product development. Nowadays it’s fashionable to call it “vibe-coding,” but in my case it’s conscious vibe-coding. I do not transfer the entire project to neural networks, but use them to automate routine, while maintaining a full understanding of each line of code, its structure and underlying architecture. Secondly, I have a comrade in arms - a tester friend who uses the panel for his daily tasks, mercilessly breaks new features in production, reports bugs and makes a huge contribution to development. It is thanks to such live testing that the instrument becomes truly usable. And, looking ahead, you too can join this process!

Read more