Bayer Dithering v0.3.1

Changelog

[0.3.1] - 2026-05-21

➕ Added

• Optional Dependencies: The taichi GPU backend is now an optional dependency ([gpu]), making the base installation extremely lightweight and universally compatible out-of-the-box.
  (9cbedae)
• Smart GPU Fallback: The CLI now features graceful degradation. If a user requests hardware acceleration (-a gpu) without the backend installed, the program safely intercepts the call, displays a friendly UI warning, and automatically reroutes processing to the CPU.
  (9aef246)

🔧 Changed

• Broader Python Support: By making taichi optional, the strict <3.14 Python limitation was inherently lifted. The library can now be installed and run via the CPU pipeline on any modern Python version (>=3.13).
  (9aef246)

⚙️ Fixed

• API & Import Safety: Implemented the Dummy Object pattern for GPUProcessor. This prevents Python from crashing with a NameError during the import stage of @ti.data_oriented decorators when the taichi backend is missing.
  (9aef246)
• PEP 639 Compliance: Updated the license declaration in pyproject.toml to use the standard SPDX identifier (license = "MIT"), resolving setuptools deprecation warnings during the build process.
  (9cbedae)

📁 Assets & Docs

• Updated README.md to include a new "Prerequisites" section explaining the CPU/GPU pipeline differences.
  (9cbedae)
• Enhanced installation instructions to reflect the new pip install BayerDithering[gpu] command.
  (23d50fb, c7cf875)

---

Full Changelog: v0.3...v0.3.1