Probability Distribution Visualizer

An interactive probability distribution visualizer with a Streamlit web interface, a typed Python API, and toolkits for fitting, Monte Carlo simulation, and statistical testing.

• Docs (GitHub Pages): https://sanskarpan.github.io/probviz/
• 16 univariate distributions (10 continuous + 6 discrete) with PDF/PMF, CDF, quantiles, sampling, and full statistics — in the web app and the API.
• Advanced modules (Python API): multivariate distributions, copulas, mixtures/GMM, distribution fitting, Monte Carlo, and statistical tests.
• Production-ready: 695-test suite, typed packaging (pyproject.toml), probviz CLI, Docker/Compose, CI with coverage gate, docs + PyPI + Docker release pipelines. See CHANGELOG.md.

Gallery

Top: Normal σ sweep · CLT convergence · Gaussian-copula ρ sweep. Bottom: Beta shape sweep · mixture separation. Regenerate with python examples/generate_media.py.

Quick start

pip install probviz
probviz app

Or from source:

git clone https://github.com/sanskarpan/probviz.git
cd probviz
pip install -r requirements.txt
streamlit run web/app.py

Open http://localhost:8501. Full guide: QUICKSTART.md · docs quickstart.

Install as a package

pip install -e ".[test]"
probviz app        # launch the UI
probviz test       # run tests
probviz version    # print version

Docker

docker build -t probviz .
docker run -p 8501:8501 probviz
# or
docker compose up --build

What's inside

Project layout and conventions: docs/architecture.md. API reference: docs/api.md (rendered on the docs site).

Python API

import numpy as np
from probviz.distributions import NormalDistribution, BinomialDistribution

normal = NormalDistribution(mu=0, sigma=1)
x = np.linspace(-4, 4, 200)
pdf, cdf = normal.pdf(x), normal.cdf(x)
samples = normal.rvs(size=1000, random_state=42)
print(normal.get_statistics())
print(normal.interval(0.95), normal.ppf(0.975))

binomial = BinomialDistribution(n=10, p=0.3)
print(binomial.pdf(np.arange(0, 11)))

from probviz.fitting import DistributionFitter
from probviz.monte_carlo import MonteCarloSimulator

fitter = DistributionFitter(samples)
print(fitter.fit_all())

rng = np.random.default_rng(42)
sim = MonteCarloSimulator(random_seed=42)
res = sim.estimate_probability(lambda: rng.normal(0, 1) > 1.0, num_samples=100_000)
print(res["probability"], res["confidence_interval"])

Known limitations (by design)

• The Streamlit app covers the 16 univariate distributions only; advanced modules are Python-API only.
• Clayton/Gumbel copula pdf/rvs are bivariate-only (explicit error otherwise).
• StudentTCopula has no closed-form cdf/pdf; use Monte Carlo via rvs.
• MixtureDistribution EM assumes 1-D components; use the sklearn GMM wrappers for multivariate mixtures.
• Cauchy moments are undefined — the API surfaces nan instead of masking it.

Testing

pip install -r requirements-dev.txt
pytest tests/ -q                                  # 695 tests
pytest tests/ -q --cov=src --cov-report=term      # with coverage (gate: 80%)
flake8 src/ tests/ --count --select=E9,F63,F7,F82 --statistics
mypy --config-file=pyproject.toml src/

Contributing

See CONTRIBUTING.md (setup, style, tests, PR checklist), CODE_OF_CONDUCT.md, and SECURITY.md.

License

MIT — see LICENSE. If you use this in research or teaching:

@software{probability_distribution_visualizer,
  title  = {Probability Distribution Visualizer},
  author = {sanskarpan},
  year   = {2026},
  url    = {https://github.com/sanskarpan/probviz}
}

Also see CITATION.cff.