Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.
TypeScript ai-sdkclaude-codeclicode-searchcodexcoding-agentscontext-retrievaldeveloper-toolsjevsemantic-searchtypescriptvercel-ai-gateway
docs: retain Astropy official solve and paired trace limits
David Zhang541b3c372ac4
diff --git a/specs/jevgrep/assets/freshness-confirmation.md b/specs/jevgrep/assets/freshness-confirmation.mdindex 669dd8c..01eba38 100644--- a/specs/jevgrep/assets/freshness-confirmation.md+++ b/specs/jevgrep/assets/freshness-confirmation.md@@ -16,6 +16,7 @@ separately and excluded; incomplete observations are subtotals, not totals.| --- | --- | ---: | ---: | --- || scikit-learn__scikit-learn-13124 | solved; not a cost win | $0.3434830 | $0.2944360 | known $0.046611348; total unknown || django__django-15629 | unresolved | $1.0695890 | $1.5055472 | known $0.158282796; total unknown |+| astropy__astropy-13579 | solved; cost win | $0.3914126 | $0.4286310 | known $0.077806470; total unknown |Scikit-learn returned five relevant files in 5,038 bytes, implementation readingleads and two test source blocks. The agent then read the splitter and tests, searched documentation,@@ -48,3 +49,13 @@ pre-patch SQLite read are retained in`/tmp/jg-freshness-v1-django-comparison/`. The needed code was delivered and read;this does not prove model variance alone caused the outcome, nor that addinginitial excerpts would prevent it.++Astropy returned 39 files in 6,381 bytes without excerpts. Two source reads and a+linear-coordinate reproduction led to the established dropped-axis fix. Focused+tests passed 41 twice; a broader local directory passed 64, skipped seven and+failed on the same expired leap-second data seen in the saved comparisons.+Official grading independently resolved the task. The patch computes a forward+transformation even when no axis is dropped, unlike the guarded comparator+patches; this adds work but no functional defect was established. Own regression+is a scalar linear case, with a manual high-level roundtrip rather than a committed+high-level test. Evidence: `/tmp/jg-freshness-v1-astropy-comparison/findings.md`.