This project explores how human intent can be linked to spatially organized digital content through an LLM-based virtual guide. Rather than treating language as a simple query interface, the system interprets natural-language preferences and connects them to content distributed across an immersive environment, so the experience reflects both what the user wants and where that content lives in space.
📝 The project is based on the conference paper:
- Iva Vasic, Hans-Georg Fill, Ramona Quattrini, and Roberto Pierdicca. 2024. LLM-Aided Museum Guide: Personalized Tours Based on User Preferences. In Extended Reality: International Conference, XR Salento 2024, Lecce, Italy, September 4–7, 2024, Proceedings, Part III. Springer-Verlag, Berlin, Heidelberg, 249–262. https://doi.org/10.1007/978-3-031-71710-9_18
Local environment values in .env should be set up before running the project.
public/pano.xmlcontains the panorama configuration.public/skin.jsandpublic/pano2vr_player.jsare generated tour files and are excluded from version control.- Large media folders such as
assets,images, andtilesshould be added separately if needed. - The pano tour is created with a commercial software tool which is why some files are not inlcuded