In brief: @SilaSoznaniyaYT_Bot automates a training centre end to end. A participant opens Telegram and sees the schedule, the library, their rank and the trainings they have completed, pays right there and gets an affiliate link with a QR code. The centre’s team sees the same thing from the other side: a CRM with finances, loyalty and the logs of every conversation with the AI assistant.
Client: the “Sila Soznaniya” training centre · Bot: https://t.me/SilaSoznaniyaYT_Bot · Category: bot + CRM
What matters most in this project?
- One entry point instead of spreadsheets and chats: schedule, payment, affiliate programme and course access all live in the bot.
- Payment straight inside Telegram through T-Bank, with webhook handling.
- A referral tree up to 20 levels deep with commission paid on the first five.
- The schedule updates itself: a GPT-powered scraper pulls the listings every night at 00:30.
- The “Estnost” AI assistant holds the conversation, and the whole thread lands in the CRM for the team to see.
- A CRM on Django Unfold with statistics, finance, loyalty and affiliate pages.
What is the project built with?
- Bot:
- aiogram 3.x with an asynchronous architecture.
- Backend and CRM:
- Django with Django Unfold.
- Database:
- shared by the bot and the admin panel, so the numbers always agree.
- Payments:
- T-Bank, an Init and Webhook pair.
- AI:
- an assistant with threads and conversation logging into the CRM, plus a GPT-powered schedule scraper.
How did it start?
Before automation the centre lived in three places at once. The schedule was edited by hand, payments came in as transfers and were reconciled against screenshots, affiliate commissions were counted in a spreadsheet. Every new training meant a day of admin work.
The affiliate programme was a pain of its own. A referral tree grows deep, commission runs across several levels, and any manual reconciliation drifts away from reality sooner or later. Arguing about percentages with a partner when the numbers live in a spreadsheet is impossible.
So everything was pulled into one system on a shared database: what the participant sees in the bot and what the administrator sees in the CRM are two sides of the same data.
What was built?
🤖 Telegram bot
- A “Trainings” section: schedule, listings, a “Next training” button
- A “Library” section
- Profile: rank, promo code, completed trainings
- “Yana”, “Care” and “About the project” sections, greetings and redirects
💳 Payment and affiliate programme
- Payment through T-Bank, an Init and Webhook pair
- An affiliate link with a QR code
- A referral tree up to 20 levels deep
- Commission logic across five levels
- Manual orders with commission and payouts
🛠 Web admin and CRM
- Management of users, events and orders
- Statistics, finance, loyalty, affiliate and manual-order pages
- Logs of conversations with the AI assistant
🧠 AI infrastructure
- The “Estnost” assistant with threads and CRM logging
- A GPT-powered schedule scraper: listings refresh daily at 00:30
What is the result?
| Metric | Value |
|---|---|
| Entry points for a participant | one, Telegram |
| Payment | T-Bank inside the bot |
| Referral tree depth | up to 20 levels |
| Commission levels | 5 |
| Listings refresh | automatic, daily at 00:30 |
| AI conversations | fully stored in the CRM |