generated: '2026-08-04'
method: probed
source: >-
  live probes of api.glia.com plus first-party client source in salemove/glia-functions-tools
  (src/lib/api.js, src/lib/errors.js)
docs: https://docs.glia.com/glia-dev/reference/errors
docs_access: gated
scope_note: >-
  Glia publishes no OpenAPI, and its reference documentation requires a Glia account login. Every
  convention below is either observed directly in an HTTP response from api.glia.com or read out
  of Glia's own published API client. Nothing here is inferred from a spec.
authentication:
  style: bearer
  header: Authorization
  format: 'Bearer <token>'
  alternate_token_types: [SessionId, AuthToken]
  detail: authentication/glia-authentication.yml
versioning:
  scheme: media-type
  header: Accept
  value: application/vnd.salemove.v1+json
  current: v1
  note: >-
    The version is negotiated through the Accept header rather than a URI path segment. The
    first-party client sends this Accept value on every request and comments that it is "Required
    by the API". One subsystem is an exception: scheduled triggers are served from a URI-versioned
    path, /api/v2/scheduled-triggers.
  uri_versioned_exceptions:
  - /api/v2/scheduled-triggers
idempotency:
  supported: true
  header: X-Idempotency-Key
  scope: mutating requests (POST, PUT, PATCH, DELETE)
  default: >-
    Enabled by default in the first-party client (idempotencyEnabled: true), which generates a key
    per request and sets the header on every mutation unless the caller supplies one.
  key_format: client-generated request identifier
  retention: not documented publicly
  source: https://github.com/salemove/glia-functions-tools/blob/main/src/lib/api.js
pagination:
  style: cursor
  request_params:
    page_size: per_page
    cursor: cursor
  response_fields:
    next_cursor: next_page_cursor
  note: >-
    The client comments that "API expects 'per_page' not 'limit'". Iteration continues while
    next_page_cursor is non-null; the log endpoint documents a maximum per_page of 1000.
  max_page_size: 1000
request_tracing:
  response_headers: [x-request-id, request-id]
  note: >-
    x-request-id is returned on every response including 401s, and the client surfaces it in
    formatted error output as "Request ID".
rate_limits:
  signaling: true
  response_headers:
  - x-rate-limit-limit
  - x-rate-limit-remaining
  - x-rate-limit-reset
  - retry-after
  status_code: 429
  reset_unit: unix seconds
  published_quotas: null
  note: >-
    Glia signals rate limits with headers and 429, and the first-party client backs off on
    retry-after. No numeric quota is published on any public page.
error_envelope:
  format: custom-json
  rfc9457: false
  shapes:
  - shape: platform
    fields: [message, type, ref, error_details]
    example_status: 401
    note: >-
      error_details maps a request element (e.g. Authorization) to a list of
      {message, type, ref}; ref deep-links into the gated error reference.
  - shape: validation
    fields: [error, status, details]
    example_status: 422
    note: details is a list of {attribute, detail} or a map of field to messages.
  detail: errors/glia-error-codes.yml
content_types:
  request: [application/json, application/json-patch+json]
  response: [application/json, text/plain]
  accept: application/vnd.salemove.v1+json
  note: >-
    Partial updates are sent as JSON Patch (application/json-patch+json) with /name, /description,
    /enabled and /schedule_pattern paths. Function source code is fetched as text/plain.
cors:
  access_control_allow_origin: '*'
  access_control_allow_credentials: true
  note: Observed on api.glia.com responses; the browser SDK calls the API directly.
regions:
  americas: https://api.glia.com
  eu: https://api.glia.eu
  note: >-
    Data residency is split US/EU and the status page tracks Americas and EU system status
    separately. Site credentials are region-bound.
site_scoping:
  parameter: site_ids[]
  header_or_query: query
  note: >-
    Collection endpoints are scoped to one or more sites; GET /queues without site_ids returns
    422 {"error":"Validation error","details":{"site_ids":["can't be blank"]}}.
resilience_expectations:
  client_defaults:
    timeout_ms: 30000
    retry: exponential backoff on 429 and 5xx
    circuit_breaker: 5 failures / 30s reset in the first-party client
  note: These are Glia client defaults, not a published server-side SLA.
x-evidence:
- {url: 'https://api.glia.com/operators', http_status: 401, fetched: '2026-08-04'}
- {url: 'https://api.glia.com/queues', http_status: 422, fetched: '2026-08-04'}
- {url: 'https://github.com/salemove/glia-functions-tools/blob/main/src/lib/api.js', http_status: 200, fetched: '2026-08-04'}