Release Notes

0.1.2_12.6.0 [2026-08-26]

Enhancements:

Updated gear description to clarify that it runs nvidia-smi to report GPU availability and driver status, and accepts an optional file input for standard Flywheel job workflows.
Added pyproject.toml with hatchling build system and project metadata.
Added fw_gear_fw_nvidia_cuda_test Python package with version detection.
Added initial test suite under tests/.

Maintenance:

Rebuilt Dockerfile using pytorch/pytorch:2.6.0-cuda12.6-cudnn9-runtime as the base image, replacing the previous nvidia/cuda:11.0.3-base image.
Upgraded CUDA_VERSION from 11.0.3 to 12.6.0.
Migrated to multi-stage Dockerfile with base, build, dev, and production stages.
Added uv as the package installer within the container.
Pinned setuptools to 78.1.1 to address vulnerabilities.
Added .dockerignore to limit build context.
Updated manifest.json to standardize classification fields, fix the gear category from converter to utility, and update the license to MIT.

Maintenance:

Added GitLab CI/CD pipeline configuration with pytest coverage and debug settings
Added pre-commit configuration with markdownlint hook for documentation quality
Updated gear naming from nvidia-cuda-test to fw-nvidia-cuda-test for consistency with Flywheel naming conventions
Updated gear label from "Nvidia CUDA Test" to "Flywheel Nvidia CUDA Test"
Updated maintainer format from support@flywheel.io to Flywheel <support@flywheel.io>
Updated gear classification suite from "Test" to "Utility"
Reorganized manifest.json fields alphabetically for better maintainability

Enhancements:

Added gear to test NVIDIA GPU availability by running nvidia-smi command
Added support for CUDA 11.0.3 with NVIDIA driver requirements for Tesla and compatible GPUs
Added optional file input to allow gear execution in Flywheel workflows
Added environment configuration for NVIDIA CUDA toolkit paths and driver capabilities

Documentation: