Rootless podman on cirunner-vps cannot copy cached actions into job
containers (symlinks: "path escapes from parent"), so any job that
landed there failed. Checkout is now plain git and Bun comes from its
install script; both are verified on cirunner-vps.
PostHog loads after the page is idle through a same-origin /ingest proxy and follows the signed-in Convex user. Plausible uses the per-site script through a /pa proxy. Sentry keeps errors and traces only. Pages load only the @spoon/ui modules they import, and client code reads public env from publicEnv.ts so zod stays out of the browser. Both analytics integrations stay off until their env is set in Infisical prod.
Five dependencies used "latest", so every upstream release changed the resolution and CI's frozen install refused the lockfile. They now use ranges around the versions already locked and deployed, so nothing changes at runtime.
Each job logs in as the read-only spoon-ci machine identity and exports the Infisical prod environment, so production values no longer live in Gitea. After pushing the images, CI asks the VPS to redeploy podman-spoon.service over the forced-command deploy key, with the key passed through env so the printed script never contains it. Installs survive a half-written runner cache, and one production deploy runs at a time.
Infisical prod carries a valid JWKS; the copy in DOTENV_PROD had escaped quotes, so the Convex env sync had been serving invalid JSON at /.well-known/jwks.json. The first sync from Infisical replaces it.
- Box init + chownInBox exec explicitly as root: under podman keep-id the
container's default user is the mapped uid, which cannot useradd/chown.
- Init renames an existing uid-1000 passwd entry (podman keep-id injects one
named after the HOST user) instead of adding a duplicate that would win the
getpwuid lookup and make whoami report the wrong name.
- Terminal connections tag their shell with a per-connection marker and reap
the process group on WebSocket close: closing the attach socket never
killed the exec'd shell, so every disconnect leaked an attached tmux client
(session stuck at the smallest zombie's size; enough dead clients starve
tmux rendering for live ones).
- Job image: nvm 0.39.5 at /usr/local/nvm (user-writable) with
/etc/profile.d/nvm.sh, matching what user dotfiles expect; docs for the box
user + ownership model.
The workspace terminal died everywhere: bun's node:http client never
emits the client 'upgrade' event (it surfaces the 101 as a plain
response), so docker-modem's exec.start({hijack:true}) waited forever —
a connected WebSocket with no bytes flowing (prod), or an instant close
when the socket was unreachable (dev). The pre-Phase-1 bridge avoided
dockerode attach for exactly this reason; the real-TTY-resize rewrite
reintroduced it.
- runtime/exec-stream.ts: speak the exec-start HTTP upgrade over a raw
net socket (docker and podman both answer 101 UPGRADED). Works under
bun and node; honors the modem's socketPath/host:port. The stream is
returned paused so the first screen paint can't race the caller's
data handler. Exec create/resize stay on dockerode (plain POSTs).
- user-container.ts: acquireUserBox re-ensures the container on every
acquire instead of trusting a cached "initialized" flag, so a box
removed behind the worker's back (manual rm, another instance's
reaper) is recreated instead of bricking terminals and agent turns
until restart.
- terminal.ts: bridge failures are logged and close the WebSocket with
a clamped human-readable reason instead of a silent catch;
xterm-session.tsx displays that reason instead of a mute
"Session ended". Also roomier terminal line spacing (1.2 -> 1.35).
The backend aiProviderProfiles authType union added 'anthropic_oauth_json'
but the Expo provider form's local AuthType type was stale, breaking
typecheck. Add the literal, a picker option, help text, and multiline
handling to match the Next app.
Claude-Session: https://claude.ai/code/session_01ScyZ1NhfN84LAnHpwhfEQk
Gabriel Brown