Evidence selection can reject the suspicion
judgments.ts returns no finding when the selected hunk is noMatch or below the location-confidence threshold. A screening probability alone is not yet a localized defect.
Inspect the supporting source ↗REPOSITORY ANALYSIS
Jev Review
Reviews a Git diff or a codebase in stages and displays review leads in a local dashboard.
E2 · source reviewedPublic material or source evidence was reviewed. A fixed-version source review requires checking a pinned revision; this is not a reproduction, performance endorsement, or safety guarantee.
- Stars
- 586
- Forks
- 14
- License
- MIT
- Last commit
- 2026-09-17
What Jev does here
Judges risk, files, evidence regions, mechanisms and severity before rule-based reviewer routing.
Connects review leads to concrete code for human inspection.
The implementation does not route only high-risk changes to a larger model. Findings are review leads, not proof of defects; compute savings were not established. README and integration source reviewed at a fixed commit; not independently run or benchmarked by this site.
What task does this project fit?
Trace a code-review concern from screening signal to a concrete source hunk.
Material analysed：
From input to action
- Noul screening identifies possible correctness, security, reliability, compatibility and test-coverage concerns.
- Choice selects supporting evidence and classifies a mechanism.
- Score evaluates severity, with conditional routing to a reviewer.
Implementation details that matter
The pipeline asks narrower questions in stages
Mechanism classification can choose noIssue; reviewer routing is conditional on severity. This reduces the claim from “the model reviewed everything” to inspectable decisions at each stage.
Inspect the supporting source ↗When to choose it
Useful when you want auditable review evidence rather than a prose-only issue list. Still run conventional tests and independently verify each retained finding.
How to inspect it yourself
Read the screening, hunk-selection and mechanism branches in judgments.ts in sequence. Follow noMatch and noIssue to their early returns before following a finding into severity and owner routing.
Limits of this review
- The review workflow was not run on a repository.
- Neither source inspection nor a severity score proves exploitability or absence of bugs.