generated: '2026-07-18'
method: searched
source: https://docs.bytez.com/http-reference/overview
docs:
- https://docs.bytez.com/model-api/docs/get-started
- https://docs.bytez.com/model-api/docs/billing
- https://docs.bytez.com/model-api/docs/understand-the-api
authentication:
  style: api-key-header
  header: Authorization
  value: 'raw Bytez key (e.g. Authorization: BYTEZ_KEY)'
  closed_source: second `provider-key` header carrying the provider account key (pass-through, never stored)
  detail: authentication/bytez-authentication.yml
idempotency:
  supported: false
  notes: >-
    No idempotency-key header or param is documented. Inference calls are
    serverless single-shot POSTs; there is no create-resource surface that would
    require idempotent retry semantics.
pagination:
  supported: false
  notes: The list endpoints (models, tasks) return full arrays with no cursor/offset paging documented.
request_response:
  base_url: https://api.bytez.com
  content_type: application/json
  envelope: '{ "error": string|null, "output": result|null }'
  streaming: >-
    `stream: true` streams text token-by-token; `json: true` streams back media
    instead of returning it inline in JSON.
  media_inputs: pass images/audio/video by `url` (recommended) or `base64`.
versioning:
  scheme: uri-path
  current: v2
  path_prefix: /models/v2
openai_compatibility:
  supported: true
  endpoints: [Chat Completions, Completions, Responses]
  notes: >-
    OpenAI-compatible endpoints let you point existing OpenAI clients at Bytez.
    Prefix a closed-source model with its provider (e.g. openai/..., anthropic/...).
    Supports streaming, tool calling, and reasoning ("thinking") on supported models.
  docs: https://docs.bytez.com/http-reference/oaiCompliant/chatCompletions
rate_limit_signal:
  detail: rate-limits/bytez-rate-limits.yml
  status_code: 429
error_envelope:
  detail: errors/bytez-problem-types.yml
lifecycle:
  detail: lifecycle/bytez-lifecycle.yml