treg-worker

treg-worker group command [options]

treg-worker is the server console script of the Treg registry (`treg.worker:main` in the tools-registry package). treg is the member-facing client. This binary is the scheduled maintainer: capacity, overflow routes, deferred-task settlement, hub checks, and the read models behind Arena, the catalog, showcase, spend, and the public site.It expects the registry server's environment (database and platform keys) and is what upstream runs from cron. It is not part of the request path. Each analytics command is a bounded pass over the audit or ledger table; the web process reads only what these jobs published. A pass that stops on its time or row cap leaves a cursor for the next run.The worker does not start a money movement. asynctasks settle may complete or release a hold the request path already opened. overflow verify and hub check do spend: verify places a cheap call, and a hub check runs as the tool's maker at step prices.

capacity sweep [--only providers] [--json]

Read balances and quotas into snapshots and the latest state. --only is a comma-separated provider list (default: all).

Seed overflow routes from local data. --live also fetches aggregator catalogs and needs keys.

Re-check routes with a cheap call. --all includes rows that are not enabled. --max-usd caps a never-verified route (default 0.02). --budget-usd stops the run before estimated spend passes the budget.

Poll due tasks and complete holds that the request path already opened. Default limit 50.

Run each live hub tool's check once, as its maker. --only is one tool id.

Fold new audit rows into the rolling Arena aggregate. Default 110 seconds; the next run resumes.

Publish content-free Web provider totals.

Resume a bounded ten-day Web observation seed.

Fold new audit rows into per-endpoint, per-day reliability stats. Default 500000 rows. Dirty-bucket repairs are capped (default 20 buckets, 60 seconds total, 10 seconds each).

Preview mutable async day buckets. --apply resets readiness and queues one full repair. Without it, the command only reads.

Check coverage of the JSON manifest saved from requeue-async-stats --apply.

Add each approved app's new calls to its weekly activity rows.

Delete public pictures that no team, person, or app still uses.

Fold new ledger settles into per-key, per-tool, per-day spend. Default 50000 rows and 60 seconds of recount; the next run resumes.

Landing-page demo: recent posts on X, then forensics. --topic is repeatable. Defaults are 60 posts and 150 likes.

Blank expired error evidence past the 14-day window. Default batch 5000.

Delete email-to-click stash rows past 90 days.

Erase what an opted-out team stored in the archive.

Repair missed directory-sync changes. Omit --org to walk every connected team.

Remove single sign-on setup after paid SSO ends.

Delete expired enterprise access evidence. Default 400 days.

A base install has no database stack. The program then prints that it needs the server extra and exits 2. Upstream's message is:

$ treg-worker needs the server extra: pip install 'treg[server]'

treg(1), python(1), uv(1), pip(1), cron(1)

Source code · Homepage · Documentation