| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| caprock_0.52.2_windows_amd64.zip | 2026-09-03 | 8.9 MB | |
| caprock_0.52.2_windows_arm64.zip | 2026-09-03 | 8.1 MB | |
| checksums.txt | 2026-09-03 | 600 Bytes | |
| caprock_0.52.2_darwin_amd64.tar.gz | 2026-09-03 | 8.8 MB | |
| caprock_0.52.2_darwin_arm64.tar.gz | 2026-09-03 | 8.3 MB | |
| caprock_0.52.2_linux_amd64.tar.gz | 2026-09-03 | 8.7 MB | |
| caprock_0.52.2_linux_arm64.tar.gz | 2026-09-03 | 8.1 MB | |
| README.md | 2026-09-03 | 1.4 kB | |
| v0.52.2 source code.tar.gz | 2026-09-03 | 6.2 MB | |
| v0.52.2 source code.zip | 2026-09-03 | 6.5 MB | |
| Totals: 10 Items | 63.6 MB | 0 | |
Fixed
- Sessions cleared before 0.52.1 still said "compacting context". 0.52.1
gave
/clearand Escape their own event kinds, but only for events recorded after it shipped — the rows already in the database kept the compact's kind, and on this machine all 18 of them were the last event of their session, which is the phrase the session card shows. So 17 sessions went on claiming a compaction that never happened, and the/clearthat started the whole investigation was still captioned "was compacting context" beside a $1,062 cost and a 95%-full context.
A migration reclassifies them. It invents nothing: events.payload is stored
verbatim, so each row already names the hook that produced it, and the rewrite
reads that rather than guessing — SessionEnd with reason=clear becomes
context.clear, other continuing reasons become session.continue, and a
genuine PreCompact is left alone. No figure moves: these rows carry no cost
and no tokens, and the totals were checked against a copy of a real 665 MB
database before and after (identical to the cent).
Sessions left active by an old /clear are deliberately not closed by
the migration. Retiring a session whose process is still running is the
failure ADR-028 exists to prevent; they age to idle
on their own and close when their process exits.