Releases: AlexanderIstomin/barry-cache
Releases · AlexanderIstomin/barry-cache
Release list
0.9.3
What's Changed
- Cq interop by @AlexanderIstomin in #1
- feat: update CQ integration to support repeated domain query paramete… by @AlexanderIstomin in #2
- fix: resolve code-span IDMAP paths in drift detection by @AlexanderIstomin in #3
- Feat/budget aware loading and benchmarks by @AlexanderIstomin in #4
- feat: enhance CQ integration with structured provenance extension and… by @AlexanderIstomin in #5
- feat(workspaces): implement workspace management and selection logic by @AlexanderIstomin in #6
- fix: revert thin agent adapters in favor of full instructions to fix … by @AlexanderIstomin in #7
- feat: enhance agent memory policy and retrieval instructions across d… by @AlexanderIstomin in #8
New Contributors
- @AlexanderIstomin made their first contribution in #1
Full Changelog: 0.6.2...0.9.3
0.6.2
0.4.0
Barry Cache v0.4.0
Barry Cache now reuses parsed context snapshots between retrieval commands, reducing repeated reads of the same context files during agent sessions.
Highlights
- Added persistent context snapshot caching in
.context-cache/context-index.json. - Reused cached context for
route,search,load,resume, andreview --json. - Added source manifest validation so cached context is refreshed automatically when context files change.
- Updated docs and generated agent instructions to describe the new cache behavior.
- Bumped package version from
0.3.2to0.4.0.
Added
-
New context cache module:
- Builds a parsed snapshot of feature packs and ADR records.
- Stores the snapshot in
.context-cache/context-index.json. - Tracks source context files and directories using
mtimeMsandsize. - Rehydrates cached feature paths relative to the active repo.
- Treats the cache as disposable, so commands still work if the cache cannot be read or written.
-
New cache regression tests:
- Verifies that Barry creates and reuses a parsed context snapshot.
- Verifies that source file changes invalidate stale cached context.
Changed
routeTask,searchContext,loadContext,resumeProject, and review model generation now read through the shared context snapshot loader.readFeaturePacksnow uses the same cached snapshot path for consistency.- README now documents that repeated retrieval commands reuse
.context-cache/context-index.json. - Generated Barry agent templates now mention the disposable parsed context index.
Improved
-
Reduced duplicate disk reads during common agent workflows such as:
barry-cache resume --task "..." barry-cache load --route "..." barry-cache search --query "..."