Skip to content

Change request: Support baselining for semantic health advisory warnings #147

Description

@lymeswold

Problem: The semantic-health pass (o2b brain doctor / o2b brain health) emits two advisory warnings - concept-gap and batch-concept-inflation - that are computed fresh on every run from the full vault history. When a vault is bulk-seeded with many preferences in a short window (as happens during initial import or migration), those warnings fire permanently: concept-gap has no time component at all, and batch-concept-inflation is anchored to the stored confirmed_at timestamps rather than to the current clock, so clock time does not age them off.

These warnings are benign (advisory only, verdict watch), but they are also immovable without deleting or editing real memory - there is no way to acknowledge a historical burst and stop it recurring in the report. To put it bluntly, we are stuck with them!

Requested enhancement: Add an optional "baseline" / "acknowledge-before" watermark (e.g. a health.silence_before date in _brain.yaml, or a small command that records one) such that any finding whose underlying entries are entirely older than the watermark is excluded from the report. The intent is to let the user declare "I've seen and accepted the state up to this date" so a one-time seed burst no longer keeps the verdict pinned at watch, while findings tied to entries after the watermark continue to surface normally. No change to how findings are detected or stored - only to which ones are surfaced.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions