Releases: Fharena/context-pack
Releases · Fharena/context-pack
v0.2.17
Changed
- Refreshed the README demo GIF to show the current first-run flow: read-only
measure, explainable source/tests routing, setup dry-run, generated packs, and clean local checkpoints. - Updated the demo generator and tests so the visual asset stays aligned with the product flow.
Verification
- Local: 51 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.17 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.16
Changed
context-pack measureandcontext-pack startnow print why selected and related areas were chosen, making first-run routing inspectable before an agent reads files.- README and Korean README terminal demos now show the explainable routing output.
Verification
- Local: 51 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.16 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.15
Changed
- Generic first-run code tasks such as
fix login timeoutnow route to inferredsourceandtestsareas when no more specific area matches, instead of falling back to overview-only context. - README and Korean README document the improved pre-setup measurement behavior.
Verification
- Local: 51 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.15 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.14
Changed
context-pack measurenow infers source/test/docs/automation areas in memory when a repo has not been set up yet, so users can preview context savings before writing.context-pack/files.- Quickstart and README now lead with read-only
measurebefore setup for direct terminal trials.
Fixed
- Scope-reduction file percentages are capped at 100% for tiny repos where Read First entries include directories or missing pre-setup docs.
Verification
- Local: 49 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.14 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.13
Added
- Added
context-pack measure, a read-only proof command that previews selected areas, scope reduction, and approximate text budget without writing a generated pack. - Added
measureto the quickstart, README terminal demo, Korean README, Codex skill guidance, and packaged plugin metadata.
Verification
- Local: 48 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.13 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.12
Fixed
- Task keyword scoring now ignores common stop words so phrases like
and,the, orwithdo not select unrelated areas or make generated packs noisy.
Changed
- First-run inferred docs areas now include usage/adoption keywords so public onboarding and adoption tasks route to documentation context more reliably.
Verification
- Local: 47 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.12 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.11
Changed
- Refreshed the terminal demo GIF to show the current setup dry-run, setup, start, text-budget, and checkpoint flow.
- Switched README demo images to a package-safe GitHub raw URL so registry pages do not rely on bundled relative assets.
Verification
- Local: 46 unit tests passed, packaged CLI smoke passed, npm dry-run passed, Python build passed, twine check passed.
- CI: main and v0.2.11 tag workflows passed on Windows/Ubuntu with Python 3.11/3.12.
v0.2.10
v0.2.9
Added
- Generated context packs now include approximate Read First and repo text-budget estimates, so users can see the initial context-size reduction.
- The pack labels the estimate clearly as chars/4 and skips binary, unreadable, ignored, and larger-than-1MB files.
v0.2.8
Changed
- context-pack setup and init now preserve existing curated manifests by default instead of auto-expanding area docs on reruns.
- First-run area inference remains automatic, with explicit --infer-areas and --no-infer-areas controls when users want to opt in or out.