Evidence is selected, not generated
Jev chooses line IDs from the two reports; code copies those lines verbatim. This constrains invented quotations but does not prove the interpretation is right.
Inspect the supporting source ↗Jev Issue Radar is a read-only dashboard for GitHub duplicate-issue triage. It retrieves likely candidates, asks Jev whether each pair is duplicate, related, distinct, or insufficiently documented, and shows selected passages from both original reports for a maintainer to review.
The app never closes issues, posts comments, or changes labels. Retrieval is currently lexical and bounded to a recent-item scan; this limitation is explicit in the UI and README.
Adds structured choices or scores to the workflow; performance and cost benefits have not been independently verified.
Review scope is the pinned README protocol: lexical candidates, selected evidence-line IDs and failed comparisons. This site did not scan GitHub or rerun the author’s four-pair integration test; constrained quotations do not prove duplicate recall.
Help maintainers review potential duplicate GitHub issues without modifying the repository.
Material analysed:
Jev chooses line IDs from the two reports; code copies those lines verbatim. This constrains invented quotations but does not prove the interpretation is right.
Inspect the supporting source ↗Lexical retrieval can miss cross-language or differently worded duplicates before Jev sees them. Provider errors remain failed comparisons, not distinct verdicts.
Inspect the supporting source ↗Choose this for a review queue where a maintainer can inspect source passages. It is not an automatic issue-closing agent or proof of high duplicate recall.
Inspect lib/jev.mjs for the alpha Decisions endpoint and max-call controls in the README. Compare the hand-authored demo with the separately attributed 2/4-label integration result.