Added a rattler-build recipe for NanoVirDx using the hardcoded version read from pyproject.toml through PACKAGE_VERSION.
Declared the Linux runtime package dependencies, including the Histo, DiffSim, vxData, and direct Python imports used by NanoVirDx.
Added a manual GitHub Actions release workflow that uploads only to the internal virdx Artifactory conda channel.
Validated the workflow with actionlint, the recipe with rattler-build --render-only, and both files with git diff --check.
Lessons Learned: Pitfalls
The single-output recipe schema uses package, not the recipe key used by multi-output recipes such as vxdata-sdk. Copying that shape directly fails recipe parsing.
A local --with-solve render panicked inside rattler-build while processing channels; syntax-only rendering succeeded, so this did not provide a dependency-solve result.
Lessons Learned: Improvements
The packaging SOP could explicitly contrast the top-level shape for single-output (package) and multi-output (recipe plus outputs) recipes.
Document a standard dependency-solve validation command that works on macOS for Linux-only/noarch packages and internal channels.
