install-geoai— agent skill

Arguments: $@

Step 1 -- Check if geoai is importable

python3 -c "import geoai; print(f'geoai v{geoai.__version__}')"

• Success -> report the version and continue to Step 2 if --check or --extras was passed, otherwise stop.
• ImportError -> go to Step 3.

Step 2 -- Check optional dependencies

If --check is present in $@, run a comprehensive dependency check:

python3 -c "
deps = [
    'geoai', 'geopandas', 'rasterio', 'rioxarray', 'shapely',
    'leafmap', 'numpy', 'pandas', 'matplotlib',
]
for dep in deps:
    try:
        mod = __import__(dep)
        ver = getattr(mod, '__version__', 'unknown')
        print(f'{dep}: {ver}')
    except ImportError:
        print(f'{dep}: NOT INSTALLED')
"

If --extras is present in $@, also check deep learning dependencies:

python3 -c "
import sys
dl_deps = ['torch', 'torchvision', 'transformers', 'timm', 'segmentation_models_pytorch']
for dep in dl_deps:
    try:
        mod = __import__(dep)
        ver = getattr(mod, '__version__', 'unknown')
        print(f'{dep}: {ver}')
    except ImportError:
        print(f'{dep}: NOT INSTALLED')

try:
    import torch
    if torch.cuda.is_available():
        print(f'CUDA: {torch.version.cuda} (device: {torch.cuda.get_device_name(0)})')
    else:
        print('CUDA: not available (CPU only)')
except ImportError:
    print('CUDA: torch not installed')
"

Report the results and note any missing packages.

Step 3 -- Installation instructions

If geoai is not installed, tell the user:

geoai is not installed. Install it with:

>

`` pip install geoai-py ``

>

For GPU-accelerated AI models (object detection, segmentation), also install PyTorch:

>

`` pip install torch torchvision ``

>

For the full set of optional dependencies:

>

`` pip install "geoai-py[extra]" ``

Stop after showing the instructions. Do not attempt to install automatically unless the user explicitly asks.