LOA101

ADAGIA 101-2025.05
Artificial Intelligence: Introduction on challenges concerning LLM-based AI-tools development
Peter Kaczmarski and Fernand Vandamme
https://doi.org/10.57028/LOA-101-Z1085

Pages: 104
e-ISSN: 2953-1489
Access Request

Summary
In recent years there have been many breakthroughs in various areas of Artificial Intelligence computing technologies, mainly with respect to building general GPTs (Generative Pre-Trained Transformers), and their extensions. This book takes a snapshot of today’s state-of-the-art in the main areas of these developments from the perspective of available AI capabilities and their possible utilization in custom cloud-based AI- tools. In the first two parts, attention is given to online accessibility of the top-ranked Large Language Models (LLMs) from Anthropic, Google, and OpenAI, and to the assessment of their available capabilities such as streaming chat, Theory of Mind tests analysis, reasoning on images, and LLM-based information retrieval using RAG, while the third part highlights the path from creating a custom local AI-tool to its cloud deployment and further to its online use. On a more general level, the presented insights and code snippets form a reference for building more extensive Artificial Intelligence use cases using Python and/or .NET 8 (LTS)

Table of contents

About this book	4
1 LLM-based AI-tools: case studies	8
1.1 Introduction	8
1.2 LLM-based AI-tools for knowledge assistance	10
1.3 Adapting the answering style of LLMs	14
1.4 LLM-based scenario analysis and Theory of Mind 17	17
1.5 Document summarisation	19
1.6 Advanced use cases for LLMs	21
1.7 Conclusions	29
1.8 Appendix A: Developing LLM chat clients in Python	30
1.9 Appendix B: Chat test results overview	34
1.10 Appendix C: Theory of Mind test	40
1.11 Appendix D: Document summarisation tests	42
1.12 References	48
2 LLM-based text search using RAG: implementation and assessment	52
2.1 Introduction	52
2.2 RAG procedure summary	53
2.3 Implementation overview	54
2.4 Experimental results	56
2.5 Discussion of limitations and challenges	59
2.6 Conclusions	60
2.7 Appendix A: Python RAG implementation	61
2.8 Appendix B: Test document contents	64
2.9 References	69
3 Building LLM-based AI-tools for the cloud	72
3.1 Introduction	72
3.2 Implementing a Blazor OpenAI LLM Chat in C#	73
3.2.1 Initial Blazor project	73
3.2.2 Adding a Razor chat component	74
3.2.3 Implementation of the LLM chat functionality	76
3.2.4 Testing the chat function in debug mode	81
3.3 Deploying the chat application to Google Cloud Platform	83
3.3.1 Publishing a release version of the web application	83
3.3.2 Docker image creation	85
3.3.3 Importing Docker image to Google Artifact Registry	86
3.3.4 Securing the OpenAI API key in GCP Secret Manager	88
3.3.5 Deploying Docker image from GAR to GCP Cloud Run	89
3.4 User experience: a simple test	91
3.5 Conclusions	94
3.6 Appendix A: Code listing (llmchat.razor)	95
3.7 References	98
4 Summary	102
About the authors	104

Important Links

Klantendienst

Contact Details