University professor built RAG-search for university regulations using a chatbot in MAX

A university teacher, who has been writing code for more than twenty years, told Habr how he made a bot in the MAX messenger for his own needs, which over time grew into an intelligent search system (RAG) for the university regulations.

How the project started

The idea was born from everyday frustration: the author got tired of having to go to the university website every time to check his own schedule. The solution was a bot in MAX that answered the simple question "what do I have today?" with an actual class schedule.

• Author — professor at the Department of Information Technology, has been writing code for more than 20 years
• Bot launched in MAX messenger
• Initial function — answering the question "what do I have today?" with a schedule
• The team of the project grew to three people
• Final function — intelligent search (RAG) through university regulations, a document of hundreds of pages

How the schedule turned into RAG search

The bot quickly caught on among users and acquired a team of three people. With popularity came questions whose answers were simply not in the schedule: how to transfer to another program, how to return after expulsion, what happens to the scholarship and what you need to know about the dormitory. All of these topics are regulated by university regulations — a document of hundreds of pages, which is inconvenient to search through manually.

It was this need that led the team to develop an intelligent search for documents, which in professional terminology is called RAG (retrieval-augmented generation) — a technique in which the system first finds relevant fragments of the document, and then generates an answer based on them. The author directly calls what he implemented in his article "intelligent search through documents, aka RAG."

Why university regulations are a complex database for search

A regulatory document from a university of hundreds of pages is poorly suited to regular keyword search: student questions are formulated in free language ("what will happen if I don't pass the session?"), and the answer you're looking for can be formulated in legal language in a separate section of the document. This is a typical task for the RAG approach, where the system searches for relevant fragments of a document and generates an answer based on them, rather than simply returning word matches.

What this means

The story shows a practical path to introducing RAG in a small team: the technology, which is usually discussed as a corporate AI tool, here grew from a personal schedule bot into a working service that answers legally significant questions from university students.

Frequently asked questions

What is RAG in the context of this bot?

RAG (retrieval-augmented generation) is intelligent document search, in which the system finds relevant text fragments and generates an answer based on them; this is exactly how the author describes the final version of his bot.

How did the project start?

With a simple function — answering the question "what do I have today?" with an actual class schedule in the MAX messenger.

Who developed the bot?

A professor at the Department of Information Technology with more than twenty years of development experience; over time, a team of three people joined the project.