generated: '2026-07-17'
method: searched
source: https://docs.ai-coustics.com/
authentication:
  style: license-key-or-jwt
  ref: authentication/ai-coustics-authentication.yml
  detail: >-
    License (SDK) key on trusted servers; short-lived JWT bearer tokens on
    untrusted clients, minted from a backend via HTTP Basic auth to
    api.ai-coustics.io/v1/sdk/tokens.
token_refresh:
  supported: true
  mechanism: processor.update_bearer_token()
  detail: >-
    JWTs carry an exp claim (Unix seconds), ~1h TTL. Refresh proactively before
    expiry without recreating the processor or reloading the model.
idempotency:
  supported: false
  note: SDK is a real-time streaming audio processor, not a mutating REST resource API; no idempotency-key contract.
versioning:
  scheme: semver
  compatibility_matrix: https://docs.ai-coustics.com/reference/sdk/compatibility-matrix
  detail: SDK core, per-language bindings, and model file IDs are versioned and matched via a published compatibility matrix.
observability:
  opentelemetry: true
  ref: https://docs.ai-coustics.com/models/get-started/opentelemetry
  detail: SDK operational metrics can be exported to your own observability backend via OpenTelemetry.
telemetry:
  ref: https://docs.ai-coustics.com/reference/concepts/sdk-telemetry
  detail: Documents what telemetry the SDK sends and what it never sends.
audio_format:
  ref: https://docs.ai-coustics.com/reference/concepts/audio-format
  detail: Streams processed in real time; latency and audio-format concepts are documented separately.