In this session, we will delve into the dynamic relation between large language models (LLMs) and vector databases. LLMs, like ChatGPT, have the ability to make human-like text and respond intelligently to questions. But how do these models realize and keep context? The key lies in vectorization. Text is transformed into high-dimensional mathematical representations—vectors—that capture semantic meaning beyond individual words. These vectors let LLMs to efficiently store, compare, and retrieve information. By utilizing vector databases, we can execute fast similarity searches, enabling AI systems to find the most applicable information, even erstwhile it isn’t an exact match to the input. This capability is crucial for tasks specified as contextual search, advice systems, and natural language understanding. During the session utilizing Kotlin and LangChain4j, I will show how to convert text into vectors, store them in a database, and retrieve applicable information in real-time.
GeeCON 2025: Marcin Łapaj - Od słów do mądrości: Jak LLM i wektorowe bazy danych rewolucjonizują dane.
In this session, we will delve into the dynamic relation between large language models (LLMs) and vector databases. LLMs, like ChatGPT, have the ability to make human-like text and respond intelligently to questions. But how do these models realize and keep context? The key lies in vectorization. Text is transformed into high-dimensional mathematical representations—vectors—that capture semantic meaning beyond individual words. These vectors let LLMs to efficiently store, compare, and retrieve information. By utilizing vector databases, we can execute fast similarity searches, enabling AI systems to find the most applicable information, even erstwhile it isn’t an exact match to the input. This capability is crucial for tasks specified as contextual search, advice systems, and natural language understanding. During the session utilizing Kotlin and LangChain4j, I will show how to convert text into vectors, store them in a database, and retrieve applicable information in real-time.