Building on the Documentation Service work I wrote about last week, I have released an updated version of the Vaadin Docs Assistant.
Updates:
- Uses updated vector embeddings based on the work-in-progress Vaadin way documentation
- Completely re-written using Spring AI, taking advantage of the new advanced RAG functionality to compress and rewrite queries for better search performance
- Added guardrails to keep discussions on topic
- Uses gpt-4o-mini
Known issues:
- There are remaining issues with Flow/Hilla classification of some docs and you may get incorrect API code examples
- The documentation is still work-in-progress and does not yet cover everything. The embeddings are re-run each night so the assistant stays up to date.
This is a research project and does not come with any warranties or guarantees. That said, I would love for you to try it out and give feedback:
- Does it seem to give helpful answers?
- If you run into odd behavior, please share insights in this thread to help me pinpoint issues and improve the implementation.
