2. Install the Python stack

Three packages, three roles. Install what you need.

Use the benchmarks (lib)Generate and solve locally (tools)Develop or maintain (checkout)

pip install "mamut-routing-lib[cli]"          # models, checkers, BKS logic, remote archives, export + CLI
pip install "mamut-routing-lib[cli,pyvrp]"    # + the PyVRP solver wrapper
mamut-routing --version

With uv: uv add "mamut-routing-lib[cli]" in a project, or uv tool install "mamut-routing-lib[cli]" for the CLI alone.

uvx --from mamut-routing-tools mamut-tools --help     # no install: uv runs it in an ephemeral environment
pip install mamut-routing-tools                       # or install it
pip install "mamut-routing-tools[kayros]"             # + KAYROS for time-dependent instances

git clone git@github.com:ANR-MAMUT/MAMUT-routing.git && cd MAMUT-routing
git submodule update --init --recursive MAMUT-routing-lib MAMUT-routing-tools   # recursive: the nested lib is the one uv installs
uv sync                                                                          # publisher + lib + tools, editable, with the docs toolchain
uv run mamut-routing --version && uv run mamut-tools --version && uv run mamut-routing-publish --version
uv run pytest

The tutorials that follow assume this checkout (uv run ...); with a pip install, drop the uv run prefix and point --benchmarks-dir at your data.

Environment variables you may want: MAMUT_ROUTING_BENCHMARKS_ROOT (default --benchmarks-dir of the CLI and default root of the library), MAMUT_TOOLS_WORKSPACE (where the tools keep OSM extracts, instances and jobs). Full list in the reference. Next: Load an instance and check a solution.