MaintainerPulse

temporal prediction of dependency maintenance risk

The dependency network, colored by risk

Every node is a PyPI package in our universe with at least one dependency link; size is how many other packages depend on it, color is the calibrated probability of shipping no release in the next 12 months. Drag to pan, scroll to zoom, hover for details, click to open a package.

<40% 40–70% 70–90% ≥90% not scored

Time machine: replay the model's calibrated risk for every package at each historical monthly snapshot, 2018–2025 — watch abandonment spread through the dependency network. Drag the slider to scrub.

Systemic risk: high-risk packages others depend on

Risk ≥ 70% and at least 3 dependents inside the universe. These are the quiet single points of failure.

package	repo	dependents	risk
aiofiles	Tinche/aiofiles	34	83%	● high
toml	uiri/toml	34	92%	▲ critical
pycryptodome	Legrandin/pycryptodome	29	84%	● high
defusedxml	tiran/defusedxml	28	77%	● high
requests-toolbelt	requests/toolbelt	23	77%	● high
backoff	litl/backoff	18	84%	● high
async-timeout	aio-libs/async-timeout	17	77%	● high
mypy-extensions	python/mypy_extensions	17	82%	● high
requests-mock	jamielennox/requests-mock	17	82%	● high
parameterized	wolever/parameterized	17	86%	● high
pytest-runner	pytest-dev/pytest-runner	16	72%	● high
httpcore	encode/httpcore	14	77%	● high
toolz	pytoolz/toolz	13	72%	● high
future	PythonCharmers/python-future	13	86%	● high
nest-asyncio	erdewit/nest_asyncio	12	86%	● high
pysocks	Anorov/PySocks	11	92%	▲ critical
h11	python-hyper/h11	11	77%	● high
deprecation	briancurtin/deprecation	10	98%	▲ critical
bump2version	c4urself/bump2version	10	92%	▲ critical
cached-property	pydanny/cached-property	9	86%	● high