FunGen for Linux
================

Run it:
  ./run.sh            (recommended)
  ./fungen            (also works - the binary has an rpath to ./lib)

File associations (.funscript / .fungen):
  ./install.sh        registers FunGen as the default opener for
                      .funscript and .fungen files (per-user, no root).
  ./uninstall.sh      undoes it.
  Keep this folder where it is after running install.sh - the desktop
  entry points at the binary in place. Move it and re-run install.sh.

GPU acceleration (NVIDIA):
  FunGen ships the ONNX Runtime CUDA provider plugins beside the binary, but
  they load the CUDA 12 + cuDNN 9 runtime at run time. For GPU you need:
    - a recent NVIDIA driver
    - CUDA 12.x runtime
    - cuDNN 9.x

  Easiest path (no system install):
    ./install-cuda-runtime.sh   downloads NVIDIA's redistributable runtime
                                libraries into ./lib. It does not touch your
                                driver or install a system-wide CUDA toolkit.
                                Needs python3 + network. Run it once.

  Or install CUDA 12 + cuDNN 9 system-wide via your distro / NVIDIA.

  Without either, AI tracking/generation runs on CPU (slower but fully
  functional). The "FunGen AI" pill in the status bar shows the engine
  actually in use (CUDA / CPU).

Everything is self-contained in this folder; nothing is installed system-wide.