# Telemetry

Luxonis Hub and Hub AI use anonymous product telemetry to understand how adoption and model-conversion workflows are used and how
reliably they complete. Aggregated usage and outcome data helps us identify failure points, detect regressions, and prioritize
improvements to product flows and documentation. This page groups related measurements rather than listing every property.

## What is collected

### Luxonis Hub

 * Device adoption journey: reaching an adoption entry point, starting a concrete flow, selecting an adoption path, completing
   safe setup choices, viewing setup instructions, backend adoption attempts, quota blocks, and the authoritative adoption result.
 * Adoption flow context: a synthetic, per-flow ID and controlled metadata for the flow variant, UI or service surface, stage,
   adoption path, configuration method, and non-identifying outcome or failure category.
 * Excluded adoption data: Hub telemetry does not include user, team, or account IDs; device IDs, serial numbers, or MAC
   addresses; adoption keys or QR payloads; command text; or network configuration values.

### Hub AI

 * Model conversion journey: entering a conversion surface, submitting a safe conversion configuration, backend upload attempts
   and outcomes, conversion attempts and authoritative results, and the first later completed download when it can be safely
   observed.
 * Conversion context: the conversion flow source, input model family and format, target platform, quantization data type, target
   precision, relevant SNPE version, and YOLO version. These are controlled categories, not model identifiers or contents.
 * Flow correlation: a synthetic, per-conversion flow ID connects frontend, API, SDK, backend, and DepthAI milestones without
   using user, team, account, model, job, device, or installation identifiers. A download that cannot be linked to an originating
   conversion is recorded as a standalone anonymous event.
 * Conversion outcomes: controlled upload and conversion success or failure states, with coarse failure categories such as
   validation, unsupported configuration or model type, upload, backend, download resolution, or timeout failures.
 * Excluded conversion data: Hub AI telemetry does not include user, team, or account IDs; model or export-job IDs; filenames,
   paths, URLs, model contents, raw arguments, commands, free text, API keys, or tokens.

## What Is Not Collected

Telemetry is limited to predefined, structured fields. It does not collect:

 * Personally identifiable information (PII): names, email addresses, or user, team, and account IDs.
 * Credentials: passwords, API keys, access tokens, adoption keys, QR payloads, or environment-variable values.
 * Raw media or project data: images, video, sensor recordings, datasets, calibration media, model files, or model contents.
 * Free-form user content: model or job names, filenames, paths, URLs, commands, raw arguments, configuration contents, logs,
   stack traces, exception text, prompts, or source code.

## Opt Out of SDK and CLI Telemetry

Set `LUXONIS_TELEMETRY_ENABLED=false` before starting the Hub AI SDK or `hubai` CLI. This disables telemetry emitted by the local
SDK or CLI process; it does not prevent data you deliberately submit to Luxonis Hub, such as a model uploaded for conversion, from
being processed as part of the requested cloud workflow.

For opt-out settings for other Luxonis software products, see the [telemetry
FAQ](https://docs.luxonis.com/software-v3/support/faq/telemetry.md).