refactor(frontend)!: clean up extracted app references
Stop pulling and installing the extracted AppLaunchpad and DBProvider frontends, and remove their unused installer TLS setup. Remove the obsolete Template-specific Docker layer and update development examples and independent app repository links.
BREAKING CHANGE: install-v2.sh no longer installs AppLaunchpad or DBProvider. Install these applications from their sealos-apps repositories. Monitoring services and Desktop integrations remain available.
Signed-off-by: cuisongliu <cuisongliu@qq.com>
fix(applaunchpad): restore HPA spec verbatim on resume to prevent double ×10 and wrong workload kind
Pause now captures the full live HPA spec and stores it in the pause
annotation. Resume re-applies that spec unchanged, eliminating:
- CPU target doubling (50% → 500%) caused by re-multiplying the already
×10 averageUtilization through json2HPA on every start.
- HPA pointing at the wrong workload kind (StatefulSet instead of
Deployment) because the restore path never passed storeList.
- Elastic apps starting at 1 replica instead of the saved minReplicas.
- Repeat-pause wiping out the restore data captured by the first pause.
All three pause/start paths (pauseApp/startApp, updateReplica PauseApp/
StartApp, and updateAppResources replicas branches) now share the same
pauseResume helpers, so manual and scheduled operations behave identically.
Legacy annotations (pre-fix {target,value} format) are still handled:
restoreHPAYaml divides the historically over-stored value by 10 and
derives the correct storeList from the live app.kind.
* fix(account): close payment watches and await worker shutdown
* style: format controller regression tests
* Change error message for payment list failure
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
* ci: run Go tests with race detector
* ci: use standard race test targets
* ci: avoid coverage during controller race tests
* test: document trusted external request targets
* ci: fold account runtime tests into unit tests
* test: always run account testcontainers tests
fix(costcenter): render region billing independently
Prevent a slow or failed region consumption request from blocking all billing cards. Show successful regions immediately, mark failed regions as unavailable, and allow retrying an individual region.
* fix(frontend): guard unknown workspace quota types
Ignore quota resources that an older frontend bundle does not recognize and make quota formatting resilient to new SDK fields. Add a regression test for pod and unknown quota types.
* build(client-sdk): bump package version to 0.1.22
While Brain (system-brain) is the topmost non-minimized window, AppDock
(collapse pill included, zIndex 1000) no longer renders, so nothing
overlays the Brain iframe. The dock returns as soon as Brain loses the
top layer, is minimized, or is closed. Other apps and the floating-button
shape are unaffected.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
The official template one-click deploy lands in Brain via the OAuth page's
openBrainTemplateDeploy branch: `/oauth?login=github&openapp=system-brain&
templateName=...&templateForm=...` becomes an autolaunch for the Brain
`/deploy` route. GitHub one-click deploy (Brain #283) needs the same entry
point, but githubRepo/autoDeploy were silently ignored.
Add the symmetric openBrainGithubDeploy branch: when openapp is system-brain
and githubRepo is present, autolaunch Brain `/deploy?githubRepo=...&
autoDeploy=1` so the Brain GitHub deployer opens with the repo pre-filled and
auto-deploys once when autoDeploy=1. Works for both the token-login and
already-logged-in paths; anything malformed falls through to normal openapp
handling.
* feat(frontend): add Rybbit analytics integration with event forwarding
Integrate self-hosted Rybbit analytics as an additional sink alongside
the existing GTM pipeline, which remains fully unchanged:
- Add RybbitScript component to @sealos/gtm, driven by the
NEXT_PUBLIC_RYBBIT_HOST / NEXT_PUBLIC_RYBBIT_SITE_ID env vars;
loads nothing when unset
- Forward all track() events to window.rybbit.event() in addition to
dataLayer; object props are JSON-serialized and truncated at 512
chars per value (Rybbit caps properties at 2KB, primitives only)
- Mount RybbitScript in desktop / applaunchpad / dbprovider / template;
dbprovider and template get an analytics script for the first time
(their track() events previously had no consumer)
- Document the env vars in desktop/.env.template
Verified: @sealos/gtm, desktop, dbprovider and template production
builds pass. The applaunchpad build failure is a pre-existing echarts
SSR issue on main, confirmed unrelated via a stash control test.
* refactor(frontend): drive Rybbit config from runtime backend config
Address review feedback: replace build-time NEXT_PUBLIC_* env vars with
runtime configuration delivered by each app's existing config channel,
symmetric with how gtmId flows today. Config changes no longer require
a rebuild and can be toggled per environment at deploy time.
- RybbitScript is now props-driven (host / siteId); empty host or
siteId disables it. No process.env reads remain
- desktop: add layoutConfig.rybbitHost / rybbitSiteId (type, client
defaults, Helm values + configmap), passed through at the mount point
- applaunchpad: add analytics.rybbit.{host,siteId} to the config
schemas (server + client), example yaml and chart configmap; read in
getInitialProps alongside the existing gtm fields
- dbprovider / template: add optional analytics.rybbit section to the
config schemas, example yamls, chart configmaps and values; read in
getInitialProps and pass to the mount point
- New analytics sections are optional in the zod schemas so existing
deployments keep working when only the image is upgraded
- Drop the NEXT_PUBLIC_RYBBIT_* entries from desktop/.env.template;
document rybbitHost/rybbitSiteId in the Helm values guides
Verified: @sealos/gtm builds; desktop production build passes;
dbprovider / template / applaunchpad compile successfully
(applaunchpad still stops at the pre-existing echarts SSR issue on
/app/detail/monitor, unchanged from main).
* chore(desktop): restore .env.template to drop unrelated diff
* fix(frontend): buffer early Rybbit events and allowlist host in provider CSP
Review follow-up for the Rybbit integration:
- Queue events fired before the Rybbit script finishes loading and
flush them from RybbitScript onLoad, mirroring the dataLayer buffer
the GTM path already has
- Append the configured rybbitHost to script-src in both CSP
annotations (nginx snippet + higress) of the dbprovider/template/
applaunchpad ingresses, so /api/script.js is not blocked when the
integration is enabled; unset values render the CSP unchanged
- Wire applaunchpad rybbitHost/rybbitSiteId through values instead of
hardcoded empty strings in the configmap, matching dbprovider and
template (the config channel was previously broken)
* fix(desktop): expose pod workspace quota
* fix(providers): support pod quota in provider quota boxes
Adapt dbprovider, template and applaunchpad QuotaBoxes to render the
new pod workspace quota item: add sourceMap color / icon entries and
missing i18n keys, aligning zh label with the shared quota-dialog.
* fix(costcenter): embed Alipay cashier page via iframe
The Alipay payment URL from TradePagePay is a signed gateway URL over
900 characters long, which cannot be scanned as a QR code. Switch to
qr_pay_mode=4 (order code mode) so Alipay renders the official cashier
page, return its auto-submitting form HTML as the payment code, and
embed it in the frontend via an iframe. QR code rendering remains for
legacy data.
* refactor(pay): restore original alipay comments
* refactor(costcenter): load Alipay cashier page via iframe src
cuisongliu