Skip to content

metrics(ledger): re-evaluation counts and per-author-class review-parity rollups #9743

Description

@JSONbored

Problem

Two questions should be answerable with standing numbers instead of ad-hoc queries: how often are verdicts re-evaluated (and why), and do PRs face the same scrutiny regardless of author class?

Scope

  • Re-evaluation series: weekly count of reason-coded re-runs by reason code, as a share of total verdicts
  • Parity rollups, computed per window from the ledger: reviews per PR, findings raised per PR, block/close rate, and hold rate — broken out by author class (maintainer-authored vs contributor-authored), with per-repo granularity
  • Definitions documented next to the code (what counts as a finding, how classes are derived from repo permissions)
  • Exposed on the public stats endpoint alongside existing fairness aggregates; included in the public verifier's cross-checks

Acceptance

Rollups reproducible by an outsider from the ledger using the published definitions; class derivation is mechanical (permissions), not a maintained list.

Metadata

Metadata

Assignees

Labels

maintainer-onlyOwner-only work — yields no Gittensor points.

Projects

No projects

Relationships

None yet

Development

No branches or pull requests

Issue actions