Skip to main content

​
Overview

This page describes the telemetry our SDK sends. We collect a strictly limited set of product telemetry for the sole purposes of measuring SDK usage, optimizing performance, and ensuring software reliability.
All audio processing happens locally on your device. We NEVER send audio content, raw samples, recordings, transcripts, or model input/output feature data to our servers.

​
What We Collect

We collect a small set of product and usage telemetry:
  • SDK version and wrapper type (e.g. Python SDK 2.1)
  • Model identifier in use (e.g. quail-vf-2.1-l-16khz)
  • Operating system and CPU architecture
  • How much audio time is processed, reported per processing session
  • Whether the VAD is being used

​
Authentication and Metering

Each request to our backend is authenticated with your SDK license key (or a minted token), the same way an Authorization header authenticates an HTTP request. The key is a credential, not analytics. We do not collect it as telemetry about you. When a processing session starts, we create a session ID that links that session’s usage to your account. This is how we meter usage and enforce plan limits. The usage telemetry above is sent as pseudonymous usage records tied to your account.

​
What We Do Not Collect

  • Audio data of any kind
  • Recordings or buffered audio
  • Transcripts or speech content
  • Model input/output feature data
This page is a technical reference and may change as the SDK evolves. The binding terms covering data and privacy are set out in our Privacy Policy and the applicable license agreement, which this page supports with technical detail.