generated: '2026-08-26'
method: searched
source: >-
  https://github.com/qcware/promethium-examples (the provider's only public API documentation reachable
  without an account) and openapi/qc-ware-promethium-openapi.yml
docs: https://app.promethium.qcware.com/docs/api
note: >-
  QC Ware publishes NO rate limits for the Promethium REST API. The OpenAPI declares no 429 response on any of
  its 13 operations and no rate-limit response headers; the public examples repository documents none; and the
  hosted API reference at https://app.promethium.qcware.com/docs/api returns 403 to an anonymous client, so any
  limits stated there could not be read. The practical throughput constraint on this API is commercial rather
  than a request quota — workflows are metered by GPU compute second (see plans/qc-ware-plans-pricing.yml) and
  the create_workflow operation requests an explicit GPU resource class and count. Recorded as an honest zero.
limits: []
limit_count: 0
headers:
  published: []
  observed: []
  note: >-
    No X-RateLimit-*, RateLimit-* or Retry-After headers are documented, and none could be observed because
    every operation requires an X-API-KEY we do not hold.
exhaustion_status_code: null
retry_guidance: not documented
quota_model:
  type: metered-compute
  unit: GPU compute hour, billed per second
  concurrency: >-
    Not capped in published material. The pricing page states that tasks running simultaneously each accrue
    compute time separately, which implies concurrent workflows are permitted and billed, not throttled.
  resource_request:
    field: resources.gpu_type
    values: [a100, v100]
    field2: resources.gpu_count
    default: 1