FOSS Applications

Immich v3.3.6 disables automatic Microsoft telemetry in ML container

Key Takeaway

Immich has disabled default anonymous metric reporting in the ONNX Runtime used by its machine learning container.

Summary

Immich has released version 3.3.6 to address a privacy regression involving the machine learning (ML) component. A recent update inadvertently enabled telemetry within the ONNX Runtime, causing the ML container to transmit anonymous metric data to Microsoft by default.

This patch specifically targets the ML subsystem to ensure that data collection is disabled. The fix was implemented following community reports that flagged the unexpected outbound communication from the self-hosted environment.

Addressing ONNX Runtime telemetry

  • ML Subsystem Fix: The update specifically modifies the machine learning container configuration to disable telemetry within the ONNX Runtime.
  • Telemetry Source: The issue originated from a change in the ONNX Runtime dependency, which began enabling anonymous metric collection by default.
  • Privacy Correction: The patch reverts the behavior to ensure that self-hosted instances do not automatically send data to external Microsoft servers.

Why it matters

For users of self-hosted photo management solutions, this update restores the privacy guarantees expected from local-first software. It prevents unexpected outbound network traffic from machine learning workloads, which is critical for sysadmins managing sensitive data in isolated or privacy-conscious environments.

Original reference: github.com →