Telegram bot that takes orders overnight

Self-initiated demo - not a client project. Cedarline Bakehouse is an invented bakery.

A bakery takes its orders through Telegram. The messages arrive at night. The owner answers in the morning — and by then some of those customers have bought their cake somewhere else.

This is a self-initiated demonstration. Cedarline Bakehouse is an invented bakery, written for this demo.

The bot finishes the order instead of taking a message. It walks the customer through the menu, the servings, the date, delivery or pickup, the address and the contact details, and ends with a confirmed order and a number — no callback needed.

It knows the bakery's rules, not just its menu. A cake needs 48 hours, a pastry platter 24, bread 12. Mondays are closed. Delivery has zones, a price and a minimum order. When a customer asks for tomorrow and tomorrow is impossible, the bot says why in plain words and offers the dates that work, instead of a silent rejection.

It understands people who do not press buttons. "25.09", "25 Sep", "tomorrow" — typed, not tapped — all land correctly. The customer can go back and change something already chosen. And when the question is not about an order at all, the bot stops guessing and passes it to a human.

The owner wakes up to work, not to a backlog. Each finished order arrives on the owner's phone the moment it is placed, and a summary of the whole night is waiting in the morning: what was ordered, for when, for how much, and what still needs a personal answer.

One night, replayed: eleven chats between 22:05 and 04:58 became eight orders and $740 booked overnight, one question passed to a human, one cancellation and one conversation the customer abandoned. Nobody waited until morning. Every one of those numbers is counted by the page as it opens — none of them is typed into the text.

You can try it yourself. The demo page opens from a folder: no installation, no account, no internet. What you try on the page is the bot itself, not a mock-up of it — a test fails if the two ever drift apart. Nothing is borrowed from other people's code, so there is nothing to keep patched, and a hiccup in the network can never turn one order into two. The bot's token lives in its own file and never reaches the code or the logs. 64 automated tests pass.

Honestly stated: the bot has not been switched on in a live Telegram yet — that needs a token the owner creates. The same code is what runs it once the token is added.

In a real project you get: your menu, prices, lead times and delivery zones inside the bot, the orders wherever you want them, and a monthly watch — menu and price updates, Telegram's own changes handled, a report of how many orders came in and where people dropped out, and backups of every order.

Try the live demo: https://flowlab-dev.github.io/demo/telegram-orders/