Install

The core

The core suite needs only pydantic and numpy, and the reference backend is pure Python, so the whole benchmark suite runs — and tests in CI — with no simulator installed at all.

pip install qnetbench

From a clone, for development:

git clone https://github.com/grageragarces/Bench-network-applications
cd Bench-network-applications
pip install -e ".[dev]"     # core + pytest, mypy, ruff

Python 3.10 or newer.

Optional extras

Extra	Install	Gives you
sequence	pip install "qnetbench[sequence]"	the SeQUeNCe backend
netsquid	see below	the NetSquid backend
mqt	pip install "qnetbench[mqt]"	loading Qiskit / MQT Bench circuits as DQC demand
viz	pip install "qnetbench[viz]"	scripts/plot_curves.py, the characterization figures
docs	pip install "qnetbench[docs]"	building this documentation site
dev	pip install "qnetbench[dev]"	pytest, mypy, ruff

NetSquid ships from its own package index and requires a free registration at netsquid.org:

pip install --extra-index-url https://pypi.netsquid.org "qnetbench[netsquid]"

One virtualenv per simulator

SeQUeNCe and NetSquid cannot coexist

SeQUeNCe pins numpy >= 2.3.5 and NetSquid pins numpy < 2. The two extras are mutually unsatisfiable in a single environment. This is inherent to the simulators, and it is precisely the fragmentation this suite exists to paper over.

Use a separate virtualenv per simulator:

python -m venv .venv      && .venv/bin/pip install -e ".[dev,sequence]"
python -m venv .venv-ns   && .venv-ns/bin/pip install --extra-index-url https://pypi.netsquid.org -e ".[dev,netsquid]"

Each environment runs the full reference test suite plus its own simulator's tests, skipping the other's. Applications are unchanged across both — that is the point of the portable API.

Verify the install

qnetbench list          # 27 core protocols + 3 policies
qnetbench run qkd       # a full run and report on the reference backend
pytest                  # physics, app invariants, trace round-trip, policies, metrics

Building these docs

pip install -e ".[docs]"
mkdocs serve            # live-reloading preview; open the URL it prints
mkdocs serve -o         # ... or have it open a browser for you
mkdocs build            # static site into site/
mkdocs gh-deploy        # publish to GitHub Pages (the gh-pages branch)

The local URL is not /

site_url points at a GitHub Pages project site, so the dev server mounts under that same path: http://127.0.0.1:8000/Bench-network-applications/. mkdocs serve prints the full address on startup — follow that.

The build runs in strict mode, so a broken cross-reference or an unresolvable docstring reference fails the build instead of shipping a dead link. The API reference itself is generated from the package's own docstrings by mkdocstrings, so it cannot drift from the code.