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: ZotGPT (UC Irvine)
providerId: zotgpt
generated: '2026-07-28'
method: searched
source:
  - https://zotgpt.uci.edu/services/gateway/
  - https://zotgpt.uci.edu/services/api/
created: '2026-07-28'
modified: '2026-07-28'
reconciled: false
tags:
  - Generative AI
  - Rate Limiting
  - Budget
  - Higher Education
description: >-
  UC Irvine publishes no numeric request-rate table for either ZotGPT API — no
  requests-per-minute, no tokens-per-minute, no burst ceiling. What it publishes
  instead is a SPEND ceiling: usage is governed by dollar-denominated credits,
  per-key spend limits, and workspace budget enforcement, tiered by affiliation
  and by whether a funding source has been validated. That is a meaningful design
  choice rather than an omission — for an internal AI utility the scarce resource
  is budget, not concurrency, and metering in dollars lets one control surface
  span models whose per-token costs differ by an order of magnitude.
notes: >-
  Recorded as absence of a published numeric limit, not as an unlimited API.
  Callers should expect provider-side limits from the upstream vendor (OpenAI,
  AWS Bedrock, Google, Azure) to pass through the gateway, and should implement
  429 backoff accordingly.
sources:
  - https://raw.githubusercontent.com/api-evangelist/zotgpt/refs/heads/main/apis.yml
responseCodes:
  throttled: 429
limits:
  - name: Gateway virtual key spend limit
    scope: key
    metric: usd-spend
    limit: per-key limit set by the key owner in the Gateway workspace
    notes: >-
      Documented as "budget enforcement and per-key spend limits available." The
      caller sets it; the workspace bounds it.
  - name: Gateway Free Tier credit
    scope: user
    metric: usd-credits
    limit: '50'
    notes: '"$50 free credits for testing," affiliation-gated.'
  - name: Gateway Standard Tier budget
    scope: account
    metric: usd-budget
    limit: higher, KFS-validated
    notes: Chargeback to a campus financial account raises the ceiling.
  - name: Gateway Advanced Research budget
    scope: account
    metric: usd-budget
    limit: high-limit, case-validated
    notes: Reserved for validated research use cases.
  - name: Deprecated Azure API credit
    scope: user
    metric: usd-credits
    limit: '200'
    notes: '"$200 in initial API credits," faculty and staff only. Being retired.'
  - name: Request rate
    scope: client
    metric: requests-per-minute
    limit: not published
    notes: >-
      No RPM/TPM figure appears anywhere in ZotGPT documentation. Upstream vendor
      limits apply through the gateway.
policies:
  - name: Backoff strategy
    description: >-
      Exponential backoff with jitter on 429; honor Retry-After where the
      upstream provider sets it.
  - name: Budget exhaustion
    description: >-
      Spend exhaustion is the primary throttle. Request a tier upgrade with a KFS
      account rather than retrying.
gaps:
  - No published requests-per-minute or tokens-per-minute ceiling.
  - No published quota-headers contract (no documented X-RateLimit-* or
    budget-remaining header).
  - No published fair-use or concurrency policy for the interactive products.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com