Qualified CEQA reviewers need a repeatable way to judge whether advisory findings are useful before the tool reaches a broader audience. A shared corpus of filing packages would create unnecessary privacy and records-management risk.
- Let a pilot coordinator create local, controlled-label evidence templates.
- Calculate precision, high-severity false-negative rate, and report-time metrics without reading a package or storing reviewer notes.
- Produce an explicit go/no-go result against the documented pilot thresholds.
- Collecting or uploading CEQA records, PDF text, project names, or contact data.
- Replacing qualified reviewer judgment or deciding legal sufficiency.
- Recruiting participants, granting permission, or validating official sources.
pilot initcreates non-overwriting CSV templates with opaque package IDs.pilot summarizeaccepts only exact, controlled columns and labels.- Formula-like cells, free text, duplicate reviews, inconsistent package timing, unknown headers, and oversized files fail closed.
- Output contains aggregate counts and thresholds only.
- A result is
goonly when precision is at least 90%, high-severity false negatives are below 5%, and median report time is under five minutes.
- At least 20 permissioned NOD/NOE packages reviewed in the pilot.
- Every activated rule receives two qualified reviewer approvals.
- Aggregate evidence can be reproduced locally from the controlled CSV files.
The tool work is complete for this slice. Pilot execution is blocked on written permission, participating organizations, qualified reviewers, and an independent manual baseline. Qualitative notes remain private to the participant; only controlled aggregate labels may enter the evidence kit.