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: Vespa
providerId: vespa-ai
created: '2026-05-25'
modified: '2026-05-25'
reconciled: true
tags:
  - Rate Limiting
  - AI Search
  - Vector Database
description: Vespa serving throughput is governed by application configuration (per-container HTTP
  threads, document API concurrency, query timeout) rather than a fixed per-key request quota. The
  values below reflect Vespa's default protection mechanisms and Vespa Cloud guidance.
sources:
  - https://docs.vespa.ai/en/performance/sizing-feeding.html
  - https://docs.vespa.ai/en/reference/document-v1-api-reference.html
  - https://cloud.vespa.ai/pricing
headers:
  retryAfter: Retry-After
responseCodes:
  throttled: 429
  serverError: 503
  timeout: 504
limits:
  - name: Query throughput
    scope: cluster
    metric: queries_per_second
    limit: scales with container cluster size
    notes: Tune feed/query container resources and search threads per query to scale.
  - name: Document API throughput
    scope: cluster
    metric: writes_per_second
    limit: scales with content cluster size
    notes: Vespa Cloud customers typically achieve tens of thousands of writes per second per
      content cluster.
  - name: In-flight document operations
    scope: container
    metric: concurrent
    limit: bounded by container thread pool
    notes: Returns HTTP 429 when the per-container in-flight limit is exceeded.
  - name: Default query timeout
    scope: request
    metric: seconds
    limit: 500
    timeFrame: request
    notes: Default query timeout is 500ms; configurable per request via the `timeout` parameter.
  - name: Default feed timeout
    scope: request
    metric: seconds
    limit: 180
    timeFrame: request
    notes: Default feed timeout is 180s; configurable per request.
  - name: Visit batch
    scope: request
    metric: documents
    limit: wantedDocumentCount
    timeFrame: request
    notes: Visit operations are paginated via the `continuation` token; batch sizing is governed by
      `wantedDocumentCount`.