Skip to main content

​
Cloud usage and compute pricing contract

@akter/cloud-api describes hosted usage as commands, compute and outbound traffic. Compute replaces storage as the priced resource meter. CloudApi.billing.listPlans returns the PlanCatalog, and CloudApi.usage.get returns an organization’s Usage for a billing period. This document defines the shape of those responses. Prices and allowances come from the hosted pricing configuration and are not part of the contract.

​
Compute units

Compute is billed in compute unit-hours. The plan catalog’s computeSizes gives the unit weight of every machine size: each ComputeSize is { cpuKind, cpus, memoryMb, unitsPerHour }, and one hour of that size bills unitsPerHour compute unit-hours. cpuKind is shared or performance, cpus and memoryMb are positive integers, and unitsPerHour is greater than zero. The weights come from the hosted pricing configuration and are not fixed by the contract. computeSizes is optional so older catalogs still decode; current servers always send it.

​
Plan catalog

Each CatalogPlan carries: The compute fields are optional in the schema so that responses from servers that predate compute pricing still decode. Current servers always send them. When a compute value is present, decoding refuses it if it is negative. The command allowance, command cap, concurrent connections and command overage are unchanged.

​
Usage report

  • The runnerHours meter is measured in compute unit-hours, not machine hours. The egressGb meter is outbound traffic in decimal gigabytes.
  • UsagePricing.computeCentsPerUnitHour is the published compute overage price. It is optional for the same reason as the catalog fields, and current servers always send it.
  • A CapState with cap: "compute" reports limit and used in compute unit-hours for the current billing period.
  • Each byProject entry may carry computeUnitHours, the project’s compute for the period, and compute, an array of ComputeUsage records. Both are omitted for a project whose compute is not metered.
  • A ComputeUsage record is { environmentId, cpuKind, cpus, memoryMb, machineHours, computeUnitHours } for one machine size in one environment. environmentId is an opaque control-plane identifier. cpuKind is shared or performance. cpus and memoryMb are positive integers. machineHours holds the raw machine hours, and computeUnitHours holds the unit-hours billed for them at the size’s catalog weight. Decoding refuses negative hours, a machine with no CPU or memory, fractional CPUs or memory, an unknown CPU kind and an empty environment. It does not recompute unit-hours from the machine size, because the weights belong to the pricing configuration.

​
Deprecated storage fields

The server no longer reports storage. These fields and values stay optional or decodable so clients reading older responses keep working, and new servers emit none of them:
  • CatalogPlan.allowances.storageGb and CatalogPlan.overage.storageCentsPerGbMonth;
  • the storage-overage and storage-cap plan features;
  • the storage cap and the storageGb meter;
  • Usage.latestStorageSample, a project’s storageGbMonths, and UsagePricing.storagePerGbCents.
A storage-era response that carries none of the compute fields still decodes.