Routebook

Operator journeys

Choose the work in front of you.

Each route begins with an action and ends with something you can verify. Open the page closest to the decision you need to make, not a generic feature bucket.

1. 01

   Bring online

   Install the gateway, keep its first listener local, and make readiness a fact before any client depends on it.

   • Quick startBuild, configure, run, and smoke-test IO Gateway for the first time.
   • ConfigurationCore config.json fields, environment overrides, dashboard auth, proxy safety, and optional Codex App Server reset-credit profiles.
2. 02

   Connect capacity

   Add provider accounts, understand their state, and use the dashboard as the place to see what can serve traffic.

   • Provider accountsAdd, enable, refresh, re-auth, and test upstream provider accounts.
   • Codex reset-credit automationSafely preserve an expiring Codex reset credit through a managed local App Server profile when an account limit is reached.
   • DashboardUse the operator dashboard for accounts, testing, custom models, keys, and history.
3. 03

   Set route policy

   Draw client boundaries, choose model behavior, and decide when an account or fallback should be preferred.

   • API keysCreate managed keys with scopes, input caps, and durable first-use daily, weekly, or monthly token/request quotas.
   • Routing and modelsUnderstand gateway endpoints, model prefixes, custom aliases, failover, and routing rules.
   • Priority routingPrioritize one or more provider accounts so they are used first until unavailable or fully drained.
   • Custom modelsCreate ctm: aliases with provider targets, specific accounts, weights, and fallback chains.
4. 04

   Prove and observe

   Send a real client request, test a route deliberately, and read the account and quota evidence it leaves behind.

   • First client requestPoint a client at IO Gateway, inspect its model catalog, and prove a request reaches an upstream account.
   • Test APIValidate upstream routes in operator-bypass mode or against a selected managed-key policy.
   • Usage and quotaRead usage history, context-token charts, account health, provider quota, and useful admin endpoints.
   • NotificationsConfigure Telegram or Google Chat alerts for auth errors, upstream failures, and quota events.
5. 05

   Repair and release

   Work from the symptom outward, then use the release process only after the service is known to be healthy.

   • TroubleshootingDiagnose readiness failures, 401s, provider auth errors, custom-model misses, and notification failures.
   • DeploymentRelease with tag-triggered app deployment and push-triggered GitHub Pages deployment.