generated: '2026-07-21'
method: searched
source: https://docs.simplismart.ai/model-suite/settings/api-keys + openapi/*.yml
summary: Cross-cutting request/response semantics for the Simplismart inference and training APIs.
authentication:
  style: bearer
  scheme: http bearer (JWT)
  header: Authorization
  format: "Authorization: Bearer <API_KEY>"
  key_management: API keys are generated per organisation under Settings -> API Keys, carry an expiry (or Never), and can be deleted for immediate revocation.
  docs: https://docs.simplismart.ai/model-suite/settings/api-keys
  cross_ref: authentication/simplismart-authentication.yml
inference_compatibility:
  chat_completions: OpenAI-compatible /chat/completions surface; usable with OpenAI SDKs by pointing base_url at https://api.simplismart.live and using a Simplismart API key.
idempotency:
  supported: false
  note: No idempotency-key header or parameter is documented for the inference or training endpoints.
pagination:
  in_spec: false
  note: List endpoints (deployments, model repos, training jobs) accept offset/count via the SDK/CLI (ModelRepoListParams offset/count) but pagination is not declared in the published mini-specs.
  params: [offset, count]
request_tracing:
  header: none-in-spec
  note: The CLI supports a `--trace-id` / correlation id (auto-generated per request); the training LLM-metrics API keys off request_id.
versioning:
  scheme: none-explicit
  note: No global API version path/header; individual model endpoints are versioned in the resource name (e.g. Whisper /model/v2/infer/whisper vs /model/infer/whisper).
error_envelope:
  field: detail
  format: json
  cross_ref: errors/simplismart-problem-types.yml
rate_limiting:
  request_rate_headers: none-documented
  quotas: GPU/compute quotas (default 1x H100 + 1x L40 per org) govern training/compilation/deployment concurrency, not per-request rate limits.
  quotas_docs: https://docs.simplismart.ai/model-suite/settings/quotas
metadata:
  note: Deployments, model repos, and secrets are scoped to an organisation (org_id / ORG_ID) and optionally to Workspaces.