A sales report arrives in Telegram every morning
Data from a table, a schedule and a messenger in one scenario: yesterday's total, order count and average.
What the scenario looks like
Срабатывание по расписанию
Продажи за период▶
Отчёт▶▦
Модель▦
Отчёт в Telegram▶
Отчёт отправлен▶
- Execute
- Execute + Data
- LLM
A report you have to go somewhere to see gets looked at on Mondays. A report that arrives in a messenger gets read daily — which is why a dip is noticed on the second day rather than in the second week.
What is inside
- Scheduled trigger — the entry.
- Sales for the period reads the collection. The edge into it is a thin execution one: a find node would take an incoming value as its filter.
- The report — the model counts the total, the number of orders, the average and one line of comparison with the day before.
- Report to Telegram sends it.
The three parts are replaced independently: the collection supplies figures, the schedule decides when, Telegram decides who.
What to set up for yourself
- The collection and how sales get into it. Usually another scenario puts them there — from a till webhook or an order form.
- The prompt deliberately forbids advice. A model will happily explain a dip with the weather and seasonality having no data about either; a report with invented explanations stops being read faster than one with only numbers.
- The time in the schedule settings.
What to connect
A collection under Tables, a Telegram bot and the Schedule target. With many sales, move the date filtering into the query: it currently takes the last thousand rows, which is both expensive and imprecise at volume.