PR REVIEW LAB
Review the risky parts.
Jev classifies each hunk. A configurable gate queues the hard cases for an LLM or human.
Review setup
Closed-set onlyAdvanced thresholds
Routing thresholds
Uses the lower of Jev’s confidence and selected probability. Changes re-route existing results without another request.
Candidate decisions only. This lab creates review queues; it does not call a second LLM, approve, block, merge, or comment on GitHub.
Review decision
AWAITING CLASSIFICATIONPaste a PR to get started
Based on this snapshot only. Model scores are not proof that a change is safe or faulty.
0 of 0 hunks processed
0Skip expensive review
0LLM queue
0Human queue
0 of 0 hunks shown. Open a hunk for the original diff and every scored decision.
Inspect a diff to begin
Load the sample, paste a unified diff, or load a public GitHub PR.