Models and simulation

PharmFlux accepts a text .pfx model or a JSON model document. A small synthetic one-compartment model looks like this:

model synthetic_one_compartment
time_unit = h

parameters
  cl = 0.3 [L/h] log estimated
  v = 3 [L] log estimated

states
  central = 0 [mg]

dynamics
  d/dt(central) = (input_central - ((cl / v) * central))

outputs
  cp = (central / v) [mg/L]

dosing
  central [mg] = 1
end

Parameters, states, outputs, and doses carry explicit units. The dosing declaration makes central a valid target; input_central is the associated input in the dynamics. The full text model and equivalent JSON model are included as runnable examples.

Simulation requests use the pharmflux.run/v0.1 schema. They declare a solver, execution budgets, a regimen, sample times, tolerances, and any parameter or initial-state overrides. A regimen can include boluses, infusions, repeated doses, covariates and changes, and resets. Use the complete synthetic request as a template. observations can request a pre or post value at an event time; the returned times and sides preserve that distinction.

cargo run --release --locked -p pharmflux-cli -- run \
  conformance/models/synthetic-pbpk-24.pfx \
  conformance/requests/synthetic-pbpk-24.json > result.json

Solver names include diffsol_bdf, diffsol_tsit45, diffsol_esdirk34, diffsol_tr_bdf2, diffsol_rosenbrock23, diffsol_rodas5p, and analytic_linear_pk. Some model and operation combinations have narrower support. Select and validate the method for the model class; the runtime reports unsupported combinations rather than silently switching methods. Results include named output columns, units, sample times, event sides, and execution identity. These synthetic examples are illustrative and make no clinical prediction.

See fitting for estimation requests and browser integration for local simulation in a Web Worker.