generated: '2026-09-05'
method: searched
source: >-
  searched https://openmldb.ai/docs/en/main/ (including quickstart/sdk/rest_api.html,
  deploy/conf.html, maintain/perf_guide.html and faq/), the openaios-platform OpenAPI documents in
  openapi/, and https://github.com/4paradigm/phanthymotus — fetched 2026-09-05
name: 4Paradigm rate limits
limit_count: 0
note: >-
  No rate limits are published, and none would be meaningful: every 4Paradigm API surface in this
  record is self-hosted software the operator runs on its own infrastructure, so the only ceiling is
  the deployment's own capacity. Nothing throttles at 4Paradigm's edge because 4Paradigm operates no
  edge. There is also no rate-limit signalling to consume — no X-RateLimit-*, no RateLimit-*, no
  Retry-After, and no 429 declared in any of the 57 REST operations or 140 RPCs.
limits: []
response_headers: []
status_on_exhaustion: null
capacity_controls:
- control: Compute-unit quota
  surface: OpenAIOS Platform Billing API
  detail: >-
    Compute units are assigned per user and per group and priced (GET /computeunit/price), so an
    operator can cap consumption by account. This is a metering and chargeback mechanism, not a request
    rate limit, and the caps are set by the operator rather than published by 4Paradigm.
  evidence: openapi/4paradigm-openaios-billing.yaml
- control: Server-side thread and connection configuration
  surface: OpenMLDB
  detail: >-
    Concurrency is tuned through the deployment's own configuration (thread_pool_size, request timeouts,
    brpc settings) rather than enforced as a published quota.
  evidence: https://openmldb.ai/docs/en/main/deploy/conf.html