Skip to content

NativeRelay v0.4.0 - Stable Release

Latest

Choose a tag to compare

@VeloraTech VeloraTech released this 11 Nov 22:19
· 6 commits to main since this release

NativeRelay v0.4.0 — Stable Release

NativeRelay 0.4.0 is the stable release of the Windows collector, alongside the existing Linux collector.

Highlights

  • Windows process lifecycle monitoring using WMI when available, with Toolhelp polling fallback.
  • Scoped recursive filesystem notifications for file creation, modification, deletion, and rename.
  • Graceful JSONL shutdown handling, including Windows Ctrl+Break support.
  • Capability and health reporting so consumers can distinguish available, degraded, and unavailable collection.
  • Loss reporting for detected notification and stream drops.
  • No new runtime Python dependencies.

Platform support

  • Windows 10/11: process monitoring and scoped filesystem notifications. WMI may be unavailable on some hosts; Toolhelp polling can miss short-lived processes. File-open events are not supported. Windows 11 client and WMI-path validation remain limited.
  • Linux: existing CN_PROC and fanotify collectors remain available subject to kernel support and privileges.
  • macOS: no collector is provided.

Install

python -m pip install nativerelay==0.4.0
nativerelay --version

Validation

  • Stable tag: v0.4.0
  • GitHub Actions Windows collector validation: passed.
  • GitHub Actions Linux collector validation: passed.
  • PyPI Trusted Publishing: passed.
  • AgentTrace integration: installs nativerelay==0.4.0 from PyPI and verifies Windows process-start/process-exit and file-create/file-modify observations.

Known limitations

  • Windows file events are not attributed to the process that caused them.
  • Windows file-open events are unsupported.
  • Toolhelp polling may miss short-lived processes and does not report processes already running when collection starts.
  • WMI may be blocked by host policy or permissions.
  • Filesystem monitoring is limited to the configured scope; detected loss is reported but missed events cannot be replayed.