generated: '2026-07-19'
method: searched
source: https://leapyear-python-docs.readthedocs-hosted.com/en/4.1.1/
notes: >-
  Cross-cutting runtime semantics for LeapYear Core, derived from the published
  Python client reference (v4.1.1) and the first-party npm GraphQL packages.
  This is a self-hosted, client-library-mediated platform rather than a public
  HTTP API, so several conventions common to SaaS APIs (idempotency keys,
  request-id tracing headers, rate-limit headers, pagination params) are simply
  NOT PART OF the published contract. Absence is recorded honestly below rather
  than inferred.
access_style:
  primary: python-client
  client: leapyear.Client
  reference: https://leapyear-python-docs.readthedocs-hosted.com/en/4.1.1/api/leapyear.html
  secondary: graphql
  graphql_evidence: >-
    First-party npm packages leapyear-graphql-api (0.0.28) and
    leapyear-graphql-client (0.0.5), authored by LeapYear Technologies, Inc.,
    indicate a GraphQL API layer behind the product UI. No public GraphQL
    schema or endpoint was published.
authentication:
  style: username/password to a customer-operated server, token-backed session
  see: authentication/leapyear-technologies-authentication.yml
idempotency:
  supported: false
  evidence: >-
    No idempotency key, retry-safety header, or replay-window semantics appear
    anywhere in the published 4.1.1 reference.
pagination:
  supported: false
  evidence: >-
    No pagination parameters or cursor/response-envelope fields are documented.
    Result sizing is instead governed by differential-privacy constraints
    (see privacy_budget) and by GroupbyAggTooManyKeysError, which caps the key
    cardinality of a GroupbyAgg operation.
rate_limiting:
  supported: false
  evidence: >-
    No rate-limit headers or quota signaling documented. The functional
    equivalent is the privacy exposure budget, which is a correctness and
    confidentiality limit rather than a throughput limit.
privacy_budget:
  description: >-
    LeapYear's defining runtime semantic. Every computation draws down a
    quantified privacy exposure budget; the server interrupts computations that
    would reach or exceed the configured limit, and refuses computations over
    datasets too small to answer within the privacy profile.
  enforcement_errors:
    - HighPrivacyExposureException
    - SensitivityMultiplierTooLargeException
    - DataSetTooSmallException
    - DataSetSizeBlockedByPrivacyProfileException
  see: errors/leapyear-technologies-error-codes.yml
asynchrony:
  supported: true
  module: leapyear.jobs
  description: Long-running analyses execute as asynchronous jobs.
  errors:
    - AsyncTimeoutError
    - AsyncCancelledError
  reference: https://leapyear-python-docs.readthedocs-hosted.com/en/4.1.1/api/async.html
versioning:
  style: coupled client/server version
  description: >-
    The client and the LeapYear Core server must run matching versions; a
    mismatch is a hard error rather than a negotiated downgrade. There is no
    URL or header API version selector.
  enforcement_error: ServerVersionMismatch
  see: lifecycle/leapyear-technologies-lifecycle.yml
error_envelope:
  style: python-exception
  detail_carrier: leapyear.exceptions.ErrorInfo
  fields:
    - message
    - details
  see: errors/leapyear-technologies-error-codes.yml
authorization:
  style: grants
  module: leapyear.admin.grants
  objects:
    - Database
    - Table
    - View
    - User
  scopes: false