How-to guideApplies to: Unified engine deployments
Use read APIs to understand the instance before changing shared settings. Administrative HTTP actions use the deployment’s existing access rules.
Inspect compute
Start with execution-options, worker capacity, and worker listings. Read allocation details and assigned tasks to separate queued work from provisioning and active execution.
Deployment-profile creation and updates change shared capacity. Read the current profile revision and resource limits first. CPU is measured in millicores; memory is measured in bytes. Accepted runs can retain recorded settings while new work uses updated profiles.
Edit configuration
Read the configuration document and its revision. Prepare explicit changes, validate them, inspect diagnostics, then save with the reviewed revision. Validation and saving have separate endpoints.
Feature flags also have a revision contract. Their presence in the API does not mean every optional feature is enabled. Check capabilities and the feature’s own requirements.
Read evidence
Operation history, workload activity, audit records, metrics, and retained telemetry answer different questions. Filter by resource IDs and a bounded time range. Retained telemetry is not a promise of unlimited historical coverage.
Prometheus proxy endpoints accept their documented query values; POST can also carry URL-encoded form fields. Send query values in the URL for either method. Do not add a JSON body to that proxy.
For service settings and metrics context, see Operations. All administrative and observability operations are indexed in the API reference.Last modified on November 4, 2026
