AddRAG

Add RAG to the database you already run.

Methodology

This is a catalog with a paper trail, written for people who already operate a database.

What gets listed

An entry needs a real product and a URL we opened. The seed set was checked on 20 Nov 2026. The date on the entry is that check, not a claim that the vendor page is frozen.

Unverified

If a field was not on the page we cited, it is null in the database and shown as unverified. We do not fill it from memory, a comparison blog, or a round number that sounds right.

Pricing

Prices are quoted from the vendor page linked on the entry, including the unit they published. If the page was a calculator, a contact form, or a table we could not read cleanly, the price stays unverified.

The picker

The picker is a heuristic. Scale buckets (under 1M, about 1M to 20M, above about 20M) are ours, so the questions have stable answers. They are not a measured limit of any database. The rules prefer the database you already run when the vendor documents an in-place vector feature, and they add a dedicated store when that feature is missing, license-restricted, or the corpus is in the large bucket and latency is the priority.

History

Entry edits append to entry_history. Rows are not updated and not deleted by the admin API. The changelog reads that table.

Traffic stats

Page views are a count of HTML responses by path and UTC day. No cookie, no account, no IP stored on that table. Signup rate limits hash the connecting IP with a secret and do not store the address.