Home Max Service

Home Max Service sells repairs, cleaning and installation, passes each job to a master at a set rate and lives on the margin. Requests came from everywhere — calls, OLX, WhatsApp, Telegram, the website — and lived in people's heads. The CRM puts them in one place: a request becomes an order, the order goes to a feed, and a master takes it.

Taking an order is the one place where two people can collide, so it's a single atomic mutation in Convex: the first master to tap «Take» gets it, the second sees that it's gone. Around that sit the boring parts that make a service company run — clients, payments and refunds, payouts to masters, deadlines with escalations, an audit log of every action and a dashboard of the margin.

The AI does the reading. Calls are transcribed with Whisper and messages from every channel are parsed into a draft request — service, address, time — that an operator checks instead of typing. The main model sits behind an OpenAI-compatible gateway, with Llama on Cloudflare Workers AI as a fallback, so a provider outage doesn't stop the intake.

Masters get their own cabinet as an installable PWA and a Telegram bot; clients get notifications. It's React 19 on Cloudflare Workers with Convex behind it, and most server modules have their own tests. The live stand runs on demo data, and the code is public on GitHub.