Tweet by MarcSelwan

February 7, 2026

Did you know that it's silly easy to create AI-assisted natural language search of your Apache Iceberg catalog (in this case, R2 Data Catalog)? This was a fun experiment - using a simple Cloudflare Worker: 1. Grab the latest snapshot(s) for your given tables 1.5. (Optional), you can easily filter out fields, convert to md files, or in my case, if your namespace/table prefixes are UUIDs, create a catalog-overview file to help search know how to map the UUIDs to human readable names. 2. Upload to R2 Then in the dash, click "Next" until the end when setting up AI Search. (Ok ok,you might have to enter the name of the prefix) Profit. The default embedding model seemed to be "good enough" and I'm sure you can get better results rolling your own via AI Gateway.

Author
MarcSelwan
Date
February 7, 2026
Open on X