CmdStanPy is a lightweight, pure-Python interface to CmdStan, the command-line interface to the Stan probabilistic programming language. It provides a clean programmatic API for compiling Stan models and running all inference algorithms — NUTS-HMC MCMC, variational inference (ADVI), pathfinder, maximum likelihood estimation, Laplace approximation, and generated quantities. The library is designed for both interactive use (Jupyter notebooks) and production pipelines, with output returned as NumPy arrays, pandas DataFrames, or xarray Datasets. Its architecture separates model compilation (CmdStanModel), argument construction (Pydantic-based configs), subprocess execution (do_command), and result containers (CmdStanMCMC, CmdStanMLE, etc.), all tied together by a robust file-based workflow that leverages CmdStan's JSON config and CSV output standards. The codebase is well-structured and professionally maintained, with extensive test coverage, cross-platform CI (Linux, macOS, Windows, ARM64), and thorough documentation including a user guide and Jupyter notebook examples. It handles edge cases such as paths with spaces, missing data, and output file reconstruction. While tightly coupled to CmdStan's external executable and CSV format, the internal modules demonstrate strong engineering practices: lazy loading of draws, context managers for temporary files, singleton logging with toggle control, and optional dependency guards (xarray, polars). The repository is a mature library (v1.3.0) with active development and a clear deprecation roadmap toward v2.0.
shinystan is an R package that wraps an interactive Shiny web application for diagnosing and exploring MCMC posterior samples from Bayesian models. In production, users construct a shinystan S4 object from rstan, rstanarm, cmdstanr, or generic MCMC array/list output via as.shinystan(), then launch the GUI with launch_shinystan(). The app provides pages for exploring parameter distributions, estimating summaries, diagnosing convergence with R-hat/ESS/MCSE and HMC sampler statistics, and performing posterior predictive checks. It also exposes programmatic retrieval of summaries through retrieve() and supports deployment to shinyapps.io via deploy_shinystan().