How it works

Every sync finds new and changed session files, compresses their bytes verbatim, and keeps them in the archive you chose.

Walk the stores

The hourly job walks each supported session store and reads file metadata. A missing file is new. A source with a newer modified time, or a changed size at the same time, is ready to archive. Everything else stays untouched.

$ packbat sync

Keep the raw file

Packbat reads the source bytes, compresses them with zstd, and records their checksum. It does not normalize or redact the session. A newer source is appended as a superseding archive record, and existing archive objects are never rewritten.

Scope each machine

Each machine writes beneath its own hostname-safe name. That keeps two computers with the same harness and session id from colliding, and lets restore choose a source machine explicitly.

Derive the index

The JSONL index points from a session id to archived files, checksums, source times, and harness metadata. It is derived data beside the raw archive. The session files remain the durable record.

Catch up

The schedule runs hourly at :03. It also runs at login or wake, so a sleeping laptop catches up without waiting for the next hour. If one sync is already running, another exits without competing for the archive.