generated: '2026-09-10'
method: derived
source: openapi/modelrush-public.openapi.yaml + https://modelrush.ai/docs + live
  probes of api.modelrush.ai
conformance:
- id: openai-chat-completions-compat
  conforms: true
  domain_standard: true
  evidence: The contract itself declares compatibility -- POST /chat/completions is
    described in the OpenAPI as "an OpenAI-compatible request shape"
    (ChatCompletionRequest carries model/messages/stream/tools/tool_choice/
    parallel_tool_calls/response_format/stream_options), and the developer-tools
    repo ships a provider-neutral compatibility test harness
    (github.com/Moonveil-AI/modelrush-developer-tools/compatibility-test). The
    OpenAI Chat Completions wire shape is the de-facto domain standard for hosted
    AI inference; a client already speaking it integrates by swapping base URL and
    key.
- id: sse-streaming
  conforms: true
  evidence: POST /chat/completions declares a text/event-stream response for
    stream:true in the OpenAPI.
- id: webhook-hmac-signatures
  conforms: true
  evidence: Signed webhooks with x-modelrush-signature (t=TIMESTAMP,v1=HMAC,
    HMAC-SHA256) and x-modelrush-event-id deduplication
    (https://modelrush.ai/docs/concepts/webhooks).
- id: oauth2
  conforms: false
  evidence: The only securityScheme is http bearer with static API keys; no OAuth2
    flows are published and /.well-known/oauth-authorization-server 404s on every
    host.
- id: oidc
  conforms: false
  evidence: /.well-known/openid-configuration 404s on modelrush.ai and
    api.modelrush.ai (probed 2026-09-10).
- id: rfc9457
  conforms: false
  evidence: Errors are a custom application/json {error:{code,message,type}}
    envelope, not application/problem+json.
- id: pagination
  conforms: false
  evidence: List endpoints return complete list envelopes with no cursor or page
    parameters in the OpenAPI.
- id: idempotency
  conforms: false
  evidence: No Idempotency-Key header or replay-protection mechanism is documented
    on any write operation (see conventions/modelrush-conventions.yml).
- id: rate-limit-headers
  conforms: partial
  evidence: 429 with Retry-After is documented (docs/concepts/errors); no
    RateLimit-*/X-RateLimit-* quota headers are published.