Everything in one place
Access, activity and operations for the SLW fleet. Anything that needs you is at the top; everything else is one click down the rail.
Who's working, and on what
Loading…
Usage
Who opens which app, how often, and what they actually do once they are in. Team activity answers who is building; this answers who is using — and which access nobody is using.
People
Everyone who can reach an SLW app — the team, plus any app account that isn’t on the roster yet. Pick a person, then check the boxes: Build = develop the app in a cloud workspace (GitHub + Codespaces), Use = sign into the live app. Invites and the setup email go out for you.
Grant a bundle of apps in one move
Put someone in a group and they get its apps automatically — applied through each app's normal invite. You can still fine-tune any one person on the People section; the group is just the fast path.
Everyone outside @slw.vc
External users and pending invites, grouped by app — the team's own access lives on People.
Access log
Who opened which app, when, and from where — every app on rivent.dev, from the one door they all sit behind. This records the door opening, not what someone did once inside.
Server health
Loading…
Backups
Every job, its verified off-site copy, and an email the morning one fails.
Alert rules
Checked every 5 minutes by the same collector that feeds the Server section. Alerts email the owner once, then daily until resolved — never per-check.
How this stays honest: no job self-reports success — health is derived from the dump files on disk plus the byte-verified upload log, so a job that dies silently goes red by simply going quiet. Off-site copies live in R2 with 200-day retention.
Spend & API keys
What every app costs to run, which API keys it holds, and what would be worth doing about it — one place instead of nine provider dashboards. Approved design: mockups/spend.
Automation
Every job that runs without anyone watching — when it last did real work, and an email when one quietly stops.
Alert rules
Checked every 5 minutes by the same collector that feeds the Server and Backups sections. Alerts email the owner once, then daily until resolved — never per-check.
How this stays honest: no job self-reports health. Runner state is derived from its own queue plus what it last wrote; branch state is computed from the two branches directly, not from whether a workflow went green. The failure that prompted this page was a workflow reporting success for twelve days while doing nothing — so a green checkmark is never evidence here. Work that cannot be read shows as unknown, never as zero. Ford is measured by its own intake queue; the nightly slide recompute by its outcome — how many slides still carry yesterday's verdict.