Repository navigation
release: promote 0.1.24 to main - #91
Merged
Merged
Conversation
Keep verified Opus 5.5 compaction at medium effort with a 64000-token total unless the caller supplies a limit. Adaptive thinking has no hard thinking budget, so the old 40192 total could be exhausted before the handoff summary completed. Stream routed Anthropic v1 summarization internally while preserving the external JSON history, shared memory bounds, cancellation, and incomplete-summary rejection.
fix: correct Opus compaction budgeting in 0.1.24
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Promotes the Opus compaction fixes and version 0.1.24 from #90 to stable main. Fresh canonical Anthropic Opus 5.5 compactions use at most medium effort and a 64,000-token default total; explicit token limits remain intact. The older compact endpoint streams Anthropic generation internally while preserving the client JSON history and rejecting incomplete summaries.
The complete feature commit passed required CI. Local verification includes 120 focused tests, typecheck, privacy scan, GUI lint/i18n/tests/build, 44-page documentation build, macOS core/UI checks, and the full backend suite with only the failed legacy fixture rerun after its streaming fake was corrected. Independent Opus review identified the v1 header-timeout risk and then confirmed its fix with no remaining findings.
All six isolated 0.1.23-to-0.1.24 installed-app scenarios passed: stopped/native/owned success, owned active-Later, owned active-Update Anyway, and native active-Update Anyway. Each exercised real Sparkle replacement and relaunch, preserved runtime/integration state, and verified the new policy and compact transport in the updated bundle. Fixture apps, processes, service jobs, preferences and disposable keys were cleaned up; the actual installed app and proxy remained unchanged. The four downloaded draft assets match the canonical bytes, and the authenticated downloaded feed plus all 19 referenced archives verify. Remote tag v0.1.24 points to clean source 0d54740, whose file tree equals the development merge. Earlier harness setup and consent-driver failures are retained as evidence; the production archive was unchanged.
Release security review: the existing external Ed25519 key matches the installed trust anchor, pinned Sparkle 2.9.6 tool hashes were checked, and the authenticated feed plus every referenced archive verify locally. Historical immutable archives are preserved. No credential, destination, signing-policy, updater-trust, or release-script changes are included. The universal app remains an ad-hoc signed, unnotarized preview; both slices are verified, with live installed-app qualification on arm64.
Required promotion CI passed on development head 76bcb45. The matching push CI and macOS cross-platform verification also passed; Windows post-integration verification remains in progress and is not a PR gate.