Documents are chunked and embedded for AI retrieval. The AI assistant only answers from this knowledge base. It does not search the internet or make assumptions.

Vectorize integration is planned for a future enhancement. Currently, documents are stored in R2 and processed server-side.