AI Admission Agent

This project implements a domain-bounded conversational assistant for prospective students who need practical guidance about applying to Italian universities and handling adjacent study procedures such as student visas, residence permits, scholarships, and admission documents. The system is exposed primarily through Telegram, but its internal design is split into a Go bot layer and a Python retrieval service. The Go layer handles Telegram updates, access control, optional Redis-backed conversation memory, and optional event publication. The Python service performs guardrail checks, intent classification, retrieval-query clarification, document search, grounded answer generation, and attachment-handling logic.

The repository shows a deliberate attempt to keep the assistant narrow in scope. Queries related to study abroad in Italy are routed toward retrieval over a curated document corpus, while unrelated or unsafe requests are refused before answer generation. The corpus itself is practical rather than encyclopedic: it contains visa, residence-permit, DSU scholarship, CV, motivation-letter, and recommendation-letter material in English, Russian, and partly Kazakh, which indicates a multilingual support goal for applicants who may not be comfortable using only English or Italian.

The main contribution of the project is not a novel model, but an end-to-end engineering design that combines access-controlled messaging, document-grounded assistance, lightweight conversation memory, and operational tooling for dataset refresh and administration. At the same time, the repository also reveals important limits: the corpus is narrow, official procedures are time-sensitive, and the system depends on careful prompt design and retrieval quality to remain reliable. For that reason, the assistant is best understood as an informational support tool rather than a legal or institutional authority.