Tools for rebot : Advancing LLM-based assistants with agent integration and tool augmentation at RELEX solutions

Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis

Authors

Date

2025-07-07

Department

Major/Subject

Computer Science

Mcode

Degree programme

Master's Programme in Computer, Communication and Information Sciences

Language

en

Pages

62

Series

Abstract

Large Language Models (LLMs) have revolutionized natural language understanding and generation, yet remain limited in real-world applications due to their inability to perform complex reasoning, interact with external systems, or access real-time data. This thesis addresses these limitations by extending the capabilities of Rebot, an internal LLM-based assistant at RELEX Solutions, through the design and implementation of a centralized tools framework. By integrating the ReAct agent paradigm and enabling tool-based action execution, Rebot transitions from a passive knowledge retrieval assistant to an interactive and autonomous agent. The research employs a design and development methodology to build an extensible system architecture that supports decentralized tool development, enabling various teams across RELEX to contribute specialized tools to a centralized library. The implementation includes enhancements to Rebot’s API, response generation mechanisms, and agent-based decision-making processes. Evaluation using BLEURT and other NLP metrics demonstrates improved response accuracy and functionality, validating the efficacy of agent integration and tool augmentation. This work not only enhances operational efficiency at RELEX but also contributes a scalable blueprint for advancing LLM-based assistants in enterprise environments.

Description

Supervisor

Ylä-Jääski, Antti

Thesis advisor

Saarinen, Erkka

Keywords

large language models, ReAct agent, tool augmentation, retrieval-augmented generation, AI assistants, agent-based systems

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202508196570

Collections

[dipl] Perustieteiden korkeakoulu / SCI