Integrace datového rozhraní, jazykového modelu a komunikační platformy pro predikci sportovních výsledků

Abstract

This bachelor’s thesis focuses on the design and implementation of a Telegram bot that automatically retrieves current football statistics from the five biggest European leagues via the football-data.org API and generates predictions using the Google Gemini language model. The core of the system is a multi-level data processing pipeline: collecting statistics on teams, their head-to-head records and current form, constructing a structured prompt and passing it to the Gemini model, which generates three types of predictions: match outcome, total number of goals, and both teams to score. The application is implemented in TypeScript on Node.js using the grammY framework, a PostgreSQL database via Prisma ORM, and a three-level caching system for performance optimization. The system also includes match subscription management for sending notifications before and after matches, league table display, and support for six languages.