rateLimits: 0.1
info:
  title: Viam Rate Limits
  description: |
    Viam does not publish a hard per-key request-per-minute rate limit table. The Viam
    Platform applies fair-use and abuse-prevention quotas, with per-organization concurrency
    and bandwidth controls that scale with billing tier. Captures below summarize what is
    publicly documented or observable in viam-server defaults.
  source: https://www.viam.com/pricing
  modified: '2026-05-25'
policies:
  - id: api-fair-use
    surface: app.viam.com (AppService, DataService, MLTrainingService, BillingService)
    scope: organization
    type: fair-use
    description: |
      Cloud APIs apply fair-use limits to protect tenants. Burst behavior is tolerated but
      sustained excessive throughput will receive HTTP 429 / gRPC RESOURCE_EXHAUSTED responses.
      Contact Viam support for documented per-key quotas.
  - id: data-egress
    surface: app.viam.com (Data Client API)
    scope: organization
    type: usage-metered
    description: |
      Cloud data egress is metered at $0.25/GB on the developer plan. Sustained high-volume
      egress should use the GetDatabaseConnection API to read directly from the hot data store.
  - id: machine-grpc
    surface: viam-server (per-machine gRPC endpoint)
    scope: machine
    type: connection-limit
    description: |
      Per-machine viam-server applies session connection limits (default 100 concurrent
      sessions) and a safety-timeout heartbeat. Session timeout defaults to 2 seconds — without
      a heartbeat, viam-server stops resources to enforce safety.
  - id: ml-training-jobs
    surface: app.viam.com (MLTrainingService)
    scope: organization
    type: concurrency
    description: |
      Each organization may queue multiple training jobs; concurrent execution depends on
      cluster capacity. Contact Viam for production SLAs.
  - id: streaming-data-capture-upload
    surface: app.viam.com (DataSyncService)
    scope: machine-part
    type: bandwidth
    description: |
      StreamingDataCaptureUpload is bandwidth-limited at the per-part level to prevent a
      noisy neighbor saturating the org's ingest pipeline.
errors:
  - status: 429
    grpc: RESOURCE_EXHAUSTED
    meaning: Rate limit or quota exceeded — back off and retry with jitter.
  - status: 503
    grpc: UNAVAILABLE
    meaning: Transient unavailability — retry with exponential backoff.
guidance:
  - Use the official SDKs — they implement retry policy aligned with Viam's limits.
  - Use Data Pipelines to push aggregation to the cloud rather than polling the Data API.
  - Use GetDatabaseConnection for direct MongoDB reads when running heavy analytics.
  - For per-machine throughput, prefer WebRTC streaming over repeated unary gRPC calls.
  - Set per-organization budget alerts via the Viam Console to catch runaway costs.