Orkestia
Blog
Build apps

Operate

Watch an app in production — a public readiness probe for sign-in, continuation grants for work that outlives a session, and where logs, hosting usage and troubleshooting live.

Readiness probe

To watch your app's login without putting an organization token in a monitor, call the public readiness route with the app's client_key:

GET https://<issuer>/api/auth/end-user/readiness?client_key=<client_key>

No credential is needed. The client_key is already public in your frontend bundle.

FieldMeaning
readyTrue only when the client is configured, a seat is free, and someone is allowed to sign in
modedev or live
client_configuredRedirect URIs are set
accepting_sign_insA seat is available. A boolean, never a count
sign_in_audiencenone, allowlist, domains or any. Never the addresses
  • none means a dev app with an empty allowlist. HTTP is healthy but nobody can sign in.
  • An unknown, deleted or suspended app returns the same not-ready answer, so the route cannot be used to look up apps.
  • The route is rate limited to 60 requests per minute per caller.

Continuation grants

Agent work can outlive the session that asked for it. Instead of storing the user's JWT, the app asks for a continuation grant that can finish one named piece of work as that user.

StepWorkflowWho calls it
Issueidentity.end-user.continuation.issueThe signed-in end-user. Inputs: conversation_uuid, request_id, optional ttl_seconds (default 900, max 3600)
Redeemidentity.end-user.continuation.redeemThe organization, after the session is gone
Revokeidentity.end-user.continuation.revokeThe organization
  • The token is returned once and cannot be read again. It is single-use.
  • It is bound to the conversation and the request. A grant that names neither is refused.
  • On redeem the app, the user and the exposure are checked again. A disabled user, a suspended app or an unexposed workflow aborts the work. It never falls back to running as the organization.

Logs, metrics and failures

  • Lumen for logs, metrics and errors from your app. Its Hosting dashboard shows your sites' traffic, status codes, latency, uptime and certificate expiry.
  • Cost Center → Hosting in the console for what your sites use: edge requests, egress, storage and reserved compute. Hosting is not billed today. See Hosting usage.
  • App Host troubleshooting for sites, processes and Buzz.
  • Monitoring and debugging for failed and stuck runs.
The App Data tab in App Host has no backup or restore action today.