generated: '2026-07-20'
method: searched
source: https://docs.oxen.ai, openapi/oxen-hub-api-openapi-original.json, openapi/oxen-server-openapi-original.json
authentication:
  style: bearer-api-key
  header: 'Authorization: Bearer <OXEN_API_KEY>'
  ref: authentication/oxen-authentication.yml
base_urls:
  hub_ai_api: https://hub.oxen.ai/api/ai
  server_api: https://hub.oxen.ai/api
openai_compatibility:
  compatible: true
  note: >-
    The inference surface mirrors the OpenAI API — `POST /api/ai/chat/completions`
    and `GET /api/ai/models` are OpenAI-compatible, so existing OpenAI SDKs work by
    pointing base_url at https://hub.oxen.ai/api/ai and using an Oxen API key.
pagination:
  style: page-number
  params:
  - page_num
  - page_number
  - page_size
  note: List endpoints (generations, models, fine-tunes) page with page-number + page-size query params.
async_jobs:
  note: >-
    Long-running image/video generations are enqueued as async jobs; poll the queue
    (`GET /api/ai/queue/{generation_id}`) or generation status for completion.
  fields:
  - status  # queued | processing | completed | cancelled | error
resource_addressing:
  note: >-
    Version-control resources are addressed by namespace/repo_name plus a resource
    path that encodes branch/commit and file, e.g.
    `/api/repos/{namespace}/{repo_name}/data_frames/{resource}`.
idempotency:
  supported: false
  note: No documented Idempotency-Key header; write operations are not declared idempotent in the specs or docs.
error_envelope:
  content_type: application/json
  fields:
  - status
  - status_message
  ref: errors/oxen-problem-types.yml
rate_limiting:
  note: Requests are rate limited; exceeding limits returns a "Rate limit exceeded" error. No documented rate-limit response headers.
versioning:
  ref: lifecycle/oxen-lifecycle.yml