generated: '2026-07-21'
method: searched
source: https://docs.sailresearch.com/openapi.json
docs: https://docs.sailresearch.com/idempotency
api: openapi/sail-research-openapi-original.json
authentication:
  style: bearer
  header: "Authorization: Bearer $SAIL_API_KEY"
  format: API Key
  see: authentication/sail-research-authentication.yml
idempotency:
  supported: true
  header: Idempotency-Key
  in: header
  required: false
  max_length: 255
  scope: "keyed by (organization, API key, Idempotency-Key)"
  behavior: >-
    Retrying with the same key returns the previously stored response instead of
    re-running inference. Applies to POST /responses, /chat/completions, /messages, /batches.
  docs: https://docs.sailresearch.com/idempotency
pagination:
  style: cursor
  params: [after_id, before_id, limit]
  applies_to: [GET /batches]
metadata:
  object: metadata
  fields: [completion_window, completion_webhook, webhook_token]
  notes: >-
    Per-request metadata controls completion window (asap/priority/standard/flex latency-cost
    tier), a completion webhook URL, and a bearer token used to authenticate webhook delivery.
completion_windows:
  tiers: [asap, priority, standard, flex]
  notes: "Longer completion windows lower per-token price. See docs.sailresearch.com/completion-windows."
error_envelope:
  shape: '{ "error": { "message", "type", "param", "code" } }'
  format: openai-compatible error object
  see: errors/sail-research-problem-types.yml
versioning:
  scheme: date-based
  current: '2026-02-18'
  see: lifecycle/sail-research-lifecycle.yml
compatibility:
  openai: [Chat Completions, Responses, Models, Batches]
  anthropic: [Messages, Count Tokens]
  notes: "Drop-in migration by changing base URL to https://api.sailresearch.com/v1 and the API key."
rate_limit_signaling:
  documented: partial
  notes: "429 responses are returned (e.g. countMessageTokens); see docs.sailresearch.com/requests_at_scale for scale guidance."