Skip to content

Find code by asking what it does. A CLI for coding agents that uses Jev to discover relevant files and source context.

Record restored package runtime evidence and performance review boundaries

David Zhang58e6f6232024
diff --git a/specs/jevgrep/assets/parity-restoration.md b/specs/jevgrep/assets/parity-restoration.md
index 27af7e2..6c78a3b 100644
--- a/specs/jevgrep/assets/parity-restoration.md
+++ b/specs/jevgrep/assets/parity-restoration.md
@@ -29,10 +29,15 @@ started. No saved baseline was rerun. The frozen study and all receipts remain a
- Astropy is an operator-interrupted attempt, not a completed solve measurement.
Independent whole-spec review also found repeated same-line declaration work and
-quadratic cache-retention scanning. They remain open review items. The frozen
-reference also repeats same-line work; do not silently alter its ordinary input
-semantics while restoring parity. Cache maintenance can be improved separately
-only while preserving answer/cache behavior.
+quadratic cache-retention scanning. These remain explicit performance follow-ups,
+not fixes folded into this restoration. The frozen reference repeats same-line
+work; changing that would change its questions and require a separate experiment.
+Cache retention is finite and space-bounded, but scanning after each write can be
+slow for large caches. It does not justify changing the frozen retrieval policy
+or claiming whole-computer performance. The user's acceptance priorities remain
+solve rate and full task cost; timing is diagnostic. Any subsequent cache
+maintenance change must preserve answer/cache behavior and identify its artifact
+separately from this confirmation.
## Preservation evidence
@@ -65,7 +70,10 @@ tests, three release tests, 14 benchmark-harness tests and all 20 installed
journeys, including discovery's
declaration index, large-test suggestions, TypeScript context expansion and the
large-file anchor/selection boundary. Native and alternate-architecture
-verification still need the final restored archive.
+verification uses the final restored archive. SHA-256
+`10cc842f25a11ff1c62cb109df27cd08b9875892907f2e89eb7bac26665a1e85`
+passed native macOS arm64 on Node 24.14.0 and both Linux amd64 and arm64 installed
+smokes on Node 22.23.3. These are package/runtime checks, not live quality results.
Required product differences remain explicit: executable `jg`, Node-only runtime,
credential storage, cache, filesystem eligibility and freshness, stdout-only
diff --git a/specs/jevgrep/choices.md b/specs/jevgrep/choices.md
index 4802da0..d362111 100644
--- a/specs/jevgrep/choices.md
+++ b/specs/jevgrep/choices.md
@@ -175,3 +175,18 @@ Gap: the native smoke mechanism was delegated. Reach: portable assertions stay
shared while Linux tool-absence assertions retain their own Docker scope; a Mac
with Python installed cannot accidentally satisfy a runtime dependency through
PATH. Verdict: sound. Confidence: high.
+
+### Preserve the winner while deferring throughput changes
+
+When: reference restoration. A minified file can contain several declarations on
+one line; the winning strategy may ask about that line repeatedly under different
+declaration names. Combining those questions could change Jev's answers, so this
+port keeps them. Separately, saving a cache answer scans existing entries to
+enforce the disk limit. A large cache therefore has repeated scanning overhead.
+
+Gap: review found performance costs that are outside the user's current solve-rate
+and task-cost acceptance priorities. Reach: the confirmation does not establish
+whole-computer throughput or optimal cache maintenance. Verdict: sound for this
+confirmation, with medium confidence: retain measured retrieval behavior and
+report the cache limitation; test any later optimization as a separately identified
+artifact rather than silently changing the candidate being confirmed.