Imagine stepping into the streets of Utrecht in the 1500s. Now imagine having an all-knowing guide who can answer any question about the monuments around you.

We built a browser based VR application that gives users a 360-degree historical view of Utrecht on their phone or in a VR headset, where they can ask questions about the monuments around them. The answers are generated using Retrieval Augmented Generation, combining spatial context from the 3D environment with historical texts written by archaeologists.

In this session we will show how we built this time machine using Spring Boot, Ollama, Docling, LangChain4J, React, Typescript, cloud based LLM and LangChainJS. More importantly, we share what we learned while building the RAG pipeline: extracting text from archaeological reports, chunking sources in meaningful ways, improving retrieval quality, preventing hallucinations, and engineering context-aware prompts in the frontend.

The session concludes with a live demonstration of the application, taking us back to Utrecht in 1528 and the following centuries.