specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Snorkel AI
providerId: snorkel-ai
created: '2026-06-21'
modified: '2026-06-21'
reconciled: false
tags:
- AI
- Data Development
- Programmatic Labeling
- Weak Supervision
- SDK
- Rate Limiting
- Quotas
- Throttling
description: >-
  Snorkel Flow is an enterprise platform accessed through the snorkelflow
  Python SDK against a customer's own per-instance deployment, not a shared
  public API gateway. Snorkel does not publish public rate-limit, RPM/TPM, or
  quota documentation; throughput and concurrency are governed by the
  provisioned instance size, compute resources, and the negotiated enterprise
  agreement rather than a standardized public limit. The open-source snorkel
  library runs locally and has no network rate limits. No specific limit values
  are reconciled in this artifact.
notes: >-
  Verify any per-instance concurrency, job, or compute limits with Snorkel and
  your deployment configuration during reconciliation. There is no public
  per-account rate-limit reference to cite.
sources:
- https://docs.snorkel.ai/docs/25.3/user-guide/sdk/
- https://snorkel.ai/snorkel-flow/
responseCodes:
  throttled: 429
limits:
- name: SDK / REST Requests
  scope: instance
  metric: requests
  limit: see provider documentation
  notes: >-
    Governed by the provisioned Snorkel Flow instance and enterprise agreement;
    no public per-account RPM/TPM published.
- name: Concurrent Jobs
  scope: instance
  metric: jobs
  limit: see provider documentation
  notes: >-
    Training, labeling, and evaluation job concurrency depends on instance
    compute resources.
- name: Open-Source Library
  scope: local
  metric: usage
  limit: not applicable
  notes: Runs as a local Python dependency; no network rate limits apply.
policies:
- name: Instance-Governed Limits
  description: >-
    Throughput and concurrency follow the provisioned Snorkel Flow deployment
    and negotiated enterprise terms rather than a fixed public limit.
- name: Backoff Strategy
  description: >-
    SDK clients should implement exponential backoff with jitter and honor
    Retry-After on any 429 returned by the instance REST API.
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com