Getting Started¶
Prerequisites¶
- Linux (x86_64, glibc 2.28+) on a CPU with AVX2 (Intel Haswell / AMD Zen or newer)
- Python 3.10–3.14
Installation¶
Pre-built wheels are available for Python 3.10–3.14 on Linux x86_64. No local compilation required:
The wheels ship the ikfast and pink IK backends. The trac_ik backend
needs orocos-kdl and NLopt at build time, so it is only available in the
development setup below.
Verify installation¶
from fetch_planning.fetch import HOME_JOINTS
from fetch_planning.planning import create_planner
planner = create_planner("fetch")
goal = planner.sample_valid()
result = planner.plan(HOME_JOINTS.copy(), goal)
print(f"Planning {'succeeded' if result.success else 'failed'}")
Building Wheels from Source¶
Wheels are built with cibuildwheel inside
the manylinux Docker images (configured under [tool.cibuildwheel] in
pyproject.toml). To build one locally:
The output goes to wheelhouse/. Requirements: Docker must be installed and running.
Releasing¶
Pushing a v* tag runs .github/workflows/release.yml, which builds
manylinux_2_34 and manylinux_2_28 wheels for every supported Python and
publishes them to PyPI via Trusted Publishing:
Development Setup¶
For contributing or rebuilding C++ dependencies from source, use pixi:
git clone --recursive https://github.com/H-tr/Fetch-Planning.git
cd Fetch-Planning
bash scripts/setup.sh
Or manually: