I upgraded Cypher so it can import and search the posts I have published through Bonumark Stream on JimLunsford.net.

The import was the easy part.

The real work was getting retrieval to behave correctly. Exact searches, date ranges, source-only requests, and negative searches kept exposing problems in the routing and validation. Fixing one failure at a time was not proving the system worked. It was only proving I could patch the last thing that broke.

So I changed the test.

I restored a full backup, loaded the real 381-post archive, and built an eight-prompt acceptance suite designed to test the different ways I will actually search that content.

That gave me something individual bug fixes could not.

A repeatable standard for whether retrieval is working across the archive, not just whether one prompt happens to pass.