Omniscio

Master Sync Overseer

On a computer that develops Omniscio and keeps its master level with GitHub, the app makes sure one Master Sync overseer is always on duty. Every time it checks in, it first sees the sync's full status, measured that moment, with a RED or GREEN verdict. RED means it catches master up by hand right away. The switch is in Dev Pipeline → Setup → Master sync, and it is on by default.

What it is

  • One overseer per computer. A Master Sync crew counts as running only while one of its overseers has a live session, so a crew left behind by an overseer that died is replaced rather than mistaken for a post that is still filled. The app waits at least 30 minutes between tries and never starts a second one beside a session that is still starting — including the replacement a kept-alive overseer is itself handing its job to, which is counted from the moment it is started rather than only once it has taken the crew seat, and including the whole stretch of a hand-over taken through the crew hand-off while the outgoing overseer still shows on the roster. After three failed starts it raises one inbox card and stops trying for six hours, then tries again — a bad afternoon never ends the safety net for the life of the app.
  • The status on every check-in. The overseer's regular heartbeat and review check-ins open with the sync status: how far behind and ahead of GitHub this computer is, whether the sync has been failing and since when, when it last ran, which files clash or are still owed, and free disk space. A RED status, or one that could not be read, always reaches the overseer, even when its crew has no work open; only a GREEN status lets a quiet check-in be skipped.
  • The same rules everywhere. Every Master Sync crew loads the same standards first: catch up by hand at 50 behind or on any failing streak, merge clashes piece by piece, GitHub wins where both sides changed the same lines, never land when the import check gives no answer, start every check-in with the status, and report only numbers measured that turn.

When the sync clashes or holds a file

The overseer does not wait for its next check-in. Within about a minute of the sync recording a new clash, or holding a file whose GitHub version it could not take in, the overseer is told, with the full status. It is told once per clash and once per held file, and an app restart does not repeat it.

When RED shows

The verdict is RED when this computer is 50 or more commits behind GitHub, when the sync's newest runs keep failing (a clash, a refusal, a failed fetch), when the sync is switched off, when no run has happened for more than 30 minutes (two of the job's 15-minute runs), or when the gap cannot be measured at all.

Reading the status yourself

Run node scripts/ops/simple-master-sync.mjs --status in the Omniscio checkout. Its first lines are the verdict and the gap; --status --json prints the same facts for a script.

Cost

Each check-in is a paid agent turn. Turning the switch off starts no new overseer and leaves one that is already running alone.

Last verified 2026-11-16