feat(slop): per-PR slop score row in the maintainer panel#726
Conversation
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
gittensory-ui | 36067ec | Commit Preview URL Branch Preview URL |
Jun 14 2026, 03:40 PM |
|
Note Gittensory Gate skippedPR closed before full evaluation. No late first comment was created.
💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →. Checked by Gittensory, a quiet PR intelligence layer for OSS maintainers. |
ghost
added
the
|
gittensory · advisory review Reviewed 11 changed file(s) — two independent AI reviewers. Suggested action: ✅ Safe to merge — both reviewers found no blocking issues. Reviewer A · Suggestions
Worth double-checking
Reviewer B · Suggestions
Worth double-checking
Before = production · After = this PR's preview deploy.
|
|
Superagent didn't find any vulnerabilities or security issues in this PR. |
49461a0 to
87205b4
Compare
Completes the slop feature's visibility half: the maintainer dashboard now
shows each open PR's latest deterministic slop band + score.
- Persist slop_risk / slop_band on the cached pull_requests row (migration
0035). Written by the public-surface processor ONLY when the repo opted
into slop; gittensory-computed, so deliberately omitted from the GitHub
sync upsert's SET clause (a later sync never clobbers it).
- updatePullRequestSlopAssessment writes the assessment; the cached record
and the maintainer-dashboard reviewability rows expose slop {risk, band}
(null when unassessed).
- maintainer-panel: a new "Slop" column renders a band+score pill
(clean/low/elevated/high -> ok/info/warn/blocked tone), or an em-dash
when the repo has slop off / the PR is not yet assessed.
Tests: DB round-trip (+ latest-wins + no-op on missing row), dashboard
exposure (assessed vs null), the panel column render, and slop + AI review
coexisting in one surface pass. 97% coverage held; UI typecheck clean.
87205b4 to
36067ec
Compare
…740) Closes the outcome-learning loop by measuring whether gittensory's deterministic signals actually predict reality — without ever auto- adjusting a live score (like time-decay, recalibration stays an owner-reviewed decision). Two questions a maintainer can act on, over data that already exists: • Is the slop score predictive? For resolved PRs carrying a persisted slop band (#726), do higher-slop bands merge less often? (`discriminates`: merge rate non-increasing as band severity rises, given >=5 samples/band; null when too little resolved data.) • Are recommendations panning out? Positive (accepted/merged/improved) vs negative (rejected/closed) split from the agent recommendation-outcome ledger, scoped to the repo. - src/services/outcome-calibration.ts: pure builders (buildSlopOutcomeCalibration / buildRecommendationOutcomeCalibration / buildOutcomeCalibrationSignals) + buildRepoOutcomeCalibration env loader. - GET /v1/repos/:owner/:repo/outcome-calibration (maintainer-gated, read-only; optional ?windowDays bounds the recommendation window). No public surface, no score mutation, no reward/payout fields.
1 participant
What
Completes the slop feature's visibility half (the companion to #724's AI-advisory layer): the maintainer dashboard now shows each open PR's latest deterministic slop band + score.
How
slop_risk/slop_bandon the cachedpull_requestsrow (migration 0035). Written by the public-surface processor only when the repo opted into slop (slopGateMode != off). These are gittensory-computed signals, so they're deliberately omitted from the GitHub-sync upsert'sSETclause — a later sync never clobbers them.updatePullRequestSlopAssessmentwrites the assessment; the cached record and the/v1/app/maintainer-dashboardreviewabilityrows now exposeslop: { risk, band }(nullwhen unassessed — e.g. slop off, or PR not yet processed).maintainer-panel— a new Slop column renders a band+score pill (clean/low/elevated/high→ok/info/warn/blockedtone), or a muted em-dash when there's no assessment.Notes
z.recordOpenAPI schema, so no spec change was needed.—everywhere else.Tests
DB round-trip (latest-wins + no-op on a missing row), dashboard exposure (assessed and
nullrows), the panel column render (pill + em-dash), and slop + AI-review coexisting in one surface pass. Backend 97% coverage held (Branches 97.03%); UI typecheck + tests clean.