carlos / carloku Public

docs: the seconds-to-live deploy story (carlos deploy)

Fleet convergence is now seconds after ship+promote (platform PR #108),
and there's a one-command carlos deploy form for members with a fresh
enough CLI. Pinfra scripts stay the operator default.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Paul Campbell pushed by claude 2 months ago d7356b62a156d2408bacba894abb464da872ebe5 parent 01b5365

1 file changed, +9 −1

CLAUDE.md +9 −1

diff --git a/CLAUDE.md b/CLAUDE.md

index 3c4b036..bfc259e 100644

--- a/CLAUDE.md

+++ b/CLAUDE.md

@@ -25,4 +25,12 @@ Rules inherited from the `website` repo, which governs the sibling site:

Deploys: bucket-mode CARLOS app `carloku` on flagship — ship a clean

`git archive` export via the pinfra `scripts/ship-app.sh` / `promote-app.sh`

(see the platform repo's CLAUDE.md "Deploy reality"). Merging does NOT

-publish; the site is stale until someone ships.

+publish; the site is stale until someone ships. Since platform PR #108

+(2026-08-08), the fleet converges within seconds of ship+promote — no

+need to wait for the next adoption pass. Verify with

+`curl -sI https://carloku.com/ | grep -i x-carlos-version`, which reports

+the build the edge is actually serving. A member with a carlos CLI built

+from platform main ≥ 1fba18f can also do it in one command:

+`carlos deploy -app carloku -kind static -version <sha> <site-dir>` (ships,

+promotes, waits until serving; exit 0 means live). The pinfra scripts

+remain the operator default.