Scenario 1

Log-probability and math helpers for PyTensor-backed graphs

Scenario 2

Prior and posterior predictive simulation utilities

Scenario 3

Custom step-method blocking and proposal configuration

Scenario 4

Labeled dims and data binding via coords/Data/Minibatch

Scenario 5

Gaussian process modeling classes and HSGP approximations

Scenario 6

MCMC sampling with auto step selection (NUTS, HMC, Metropolis, Slice, Gibbs)

Scenario 7

Variational inference APIs (ADVI, SVGD, KLqp, custom approximations)

Scenario 8

Causal graph interventions/observations and model graph exports

Scenario 9

ODE-integrated probabilistic models via DifferentialEquation

Scenario 10

Model specification with distributions, deterministics, and potentials