generated: '2026-07-19'
method: searched
source: https://leapyear-python-docs.readthedocs-hosted.com/en/4.1.1/
docs: https://leapyear-python-docs.readthedocs-hosted.com/en/4.1.1/api.html
notes: >-
  LeapYear exposed its platform through a first-party Python client against a
  self-hosted server rather than a documented public HTTP contract, so these
  conventions describe the client-library surface. Idempotency, pagination, and
  request-id tracing are not documented anywhere in the public reference; those
  fields are recorded as unknown rather than assumed.
interface:
  style: client-library
  language: python
  transport: http over tls
  server: LeapYear Core (self-hosted)
  entrypoint: leapyear.Client
authentication:
  style: username/password login or public key auth, yielding a server-issued token
  see: authentication/leapyear-authentication.yml
connection:
  url_argument: url
  env_var: LY_API_URL
  default: http://localhost:4401
  lifecycle: explicit Client.close() or the Client context manager
  caching: >-
    Resource metadata is cached client-side to reduce network latency; load() on a
    resource forces a metadata refresh.
  client_flags:
  - default_analysis_caching
  - default_allow_max_budget_allocation
idempotency:
  supported: unknown
  note: No idempotency key or retry-safety contract appears in the public reference.
pagination:
  supported: unknown
  note: >-
    Collections are exposed as Python mappings and iterators (Database.all(),
    Database.tables, Database.views); no cursor or page-token contract is documented.
async:
  supported: true
  module: leapyear.jobs
  model: >-
    Long-running analyses are submitted as asynchronous jobs whose status can be
    inspected; timeouts and cancellations surface as AsyncTimeoutError and
    AsyncCancelledError.
  see: errors/leapyear-error-codes.yml
privacy_budget:
  concept: Privacy Exposure
  controls:
  - Privacy Profile assigned per database
  - Privacy Limit set per database
  - Per-computation maximum Privacy Exposure
  enforcement: >-
    Requests are refused before results are returned when the estimated exposure would
    breach the configured limit. This is the platform's defining runtime semantic and
    functions as its rate-limiting analogue: the scarce resource metered is privacy
    budget, not request volume.
  opacity: >-
    The reason a computation was blocked is deliberately not made explicit, because the
    reason could itself leak information about the underlying data.
error_envelope:
  style: python exceptions
  server_errors: leapyear.exceptions.APIError constructed from the HTTP response
  fields:
  - message
  - details
  see: errors/leapyear-error-codes.yml
versioning:
  scheme: client and server versions must match
  mismatch_error: ServerVersionMismatch
  documented_versions:
  - 4.1.0
  - 4.1.1
  see: lifecycle/leapyear-lifecycle.yml
rate_limits:
  supported: unknown
  note: No request rate limit is documented; privacy budget is the enforced quota.