generated: '2026-07-17'
method: searched
source: https://docs.adaptive-ml.com/v0.14/reference/api , https://docs.adaptive-ml.com/v0.14/advanced/authentication
description: Cross-cutting request/response semantics for the Adaptive Engine REST API.
authentication:
  style: bearer-api-key
  header: 'Authorization: Bearer <ADAPTIVE_API_KEY>'
  ref: authentication/adaptive-ml-authentication.yml
base_url:
  pattern: https://<your-adaptive-deployment>/api/v1
  note: Self-hosted; the host is the customer deployment URL (ADAPTIVE_URL), not a shared SaaS host.
openai_compatibility:
  compatible: true
  note: >-
    Chat completions and embeddings follow the OpenAI request/response shape;
    the OpenAI Python/JS clients work by pointing base_url at ADAPTIVE_URL/api/v1.
  endpoints:
  - POST /api/v1/chat/completions
  - POST /api/v1/embeddings
streaming:
  supported: true
  style: server-sent-events
  note: Chat completions support streamed deltas (ChatResponseChunk / Delta) with StreamOptions.
structured_output:
  supported: true
  field: response_format
  note: >-
    Pass plain text, a JSON Schema, or (in Python) a Pydantic model via
    response_format to constrain model output (added v0.14.0).
idempotency:
  supported: false
  note: >-
    No documented Idempotency-Key header. Large-file uploads are made
    idempotent-by-session through explicit chunked-upload session IDs
    (upload/init -> upload/part -> upload/status, with upload/abort), not via a
    generic idempotency key.
pagination:
  rest: not-documented
  note: The paginated/list surface lives in the GraphQL API and Python SDK, not the public REST OpenAPI.
error_envelope:
  format: http-status
  statuses_observed: [403, 404, 500]
  note: >-
    The REST OpenAPI returns standard HTTP status codes; responses are not
    application/problem+json (not RFC 9457). See errors/adaptive-ml-problem-types.yml.
  ref: errors/adaptive-ml-problem-types.yml
versioning:
  scheme: doc-versioned
  current: v0.14
  api_path_version: v1
  note: >-
    REST paths are /api/v1/*; the platform/SDK version (v0.14) is tracked in the
    docs URL and release notes.
  ref: lifecycle/adaptive-ml-lifecycle.yml
rate_limiting:
  signaling: not-documented
  note: Self-hosted deployment; no public rate-limit header contract documented.