DashboardAugust 2026
Three Deploy Paths & Feed Health
- Onboarding offers the CLI, a coding agent, and the AWS console as peer cards. The step used to lead with CloudFormation and bury the other two in collapsibles, so a shift toward the CLI would have read as a funnel drop rather than a preference
- Self-hosted orgs no longer see the platform CloudFormation path at all: the quick-create URL is not built, the card is not rendered, and the CLI steps and agent prompt name
wraps selfhost loginandwraps selfhost connect - Launching a stack closes the other two paths, so nobody runs
wraps email initover resources CloudFormation is still creating. The connection gate has a visible terminal state and no longer unmounts an in-flight Deploy & Connect - Retired four onboarding steps that nothing could reach: deploy-infrastructure, deploy, cli-install, and the orphaned AWS connect path
- Event-feed stall alerts judge staleness per message rather than against a cursor the webhook throttles to one write a minute. That cursor flagged every later message in a burst as unacknowledged: 13 of the 14 alerts this feature had ever sent were false, all against feeds that never missed an event
- SDK senders get a stall alert at all. Sends through
@wraps.dev/emailgo straight from your infrastructure to your SES and never touch the Wraps API, so a broken feed produced no rows and looked healthy. The hourly console-access role now reads theAWS/SESSend metric as an independent fallback, and a null probe means no evidence — never zero sends - Fix: the org webhook secret is no longer exposed through the Open AWS Console link
- Fix: AWS connection routes require
awsAccounts:write, notawsAccounts:read
Platform v0.24.0

