Sentinel — Data Model v0

Data model v0 — Sentinel by Aftergraph

Wedge guardrail: Sentinel turns pull requests into merge-ready verdicts. First wedge: CLI review on exact HEAD → SHIP / DO NOT SHIP verdict with cited evidence → GitHub App.

Entities (frozen records, immutable after issue)

Review { id, repo, prNumber, headSha, baseSha, rulePackVersion, modelId, startedAt, completedAt, status: verifying | complete | stale }
- stale is set automatically when baseSha moves mid-review. A stale review can never become a verdict.
Verdict { reviewId, decision: SHIP | DO_NOT_SHIP, findingIds[], checksPassed[], issuedAt }
- Bound 1:1 to the verified headSha. Any new push requires a new Review.
Finding { id, reviewId, severity: security | reliability | correctness | style, ruleId, file, line, evidence: string, status: open | resolved | dismissedWithReason }
- Resolved findings are never re-reported on the same HEAD.
EvidenceRef { kind: file | ciRun | diffHunk | ruleDoc, ref: string } — every finding carries ≥1.
RulePack { version, rules: [{ id, title, severity, pattern }] } — org-owned, versioned, auditable.

Explicit non-entities (v0)

No User table (GitHub identity via App; CLI runs as the invoking engineer). No auto-fix patches. No comment threads stored (GitHub owns them; we store verdict + findings only).

Source: Aftergraph/sentinel@beececfb4885d8055cc6694e8979b47759a212a4 docs/data-model-v0.md.md — verbatim, pinned, no reinterpretation.

Was this page helpful?