Catalog intelligence
Evaluation reports
Compare fast extraction against the expected synthetic seed results for all four categories.
This evaluation uses rules only, without starting the local AI. It measures development examples, not real-world equivalence or accuracy of unreviewed catalog parts.
- Report 1 — Oct. 5, 2026, 12:57 a.m.