Generate the code

Step 3 of the TVB-O spine: compile one specification into backend-native source, a study directory and a runnable workflow.

TVB-O does not interpret your specification at run time. It compiles it: the adapter for the target backend resolves everything that can be resolved in Python, and a template emits concrete, readable source — a JAX solve_fn, a TVB model class, a Julia system, a C kernel, a plot.py. That file is the deliverable. It is what you attach to a paper, and it is what a reviewer reads when they want to know what was actually integrated.

The same compilation writes the study directory the run will fill, so a spec, its generated code, its results and its figures all sit in one predictable layout.

How code generation works

Why the spec is compiled rather than interpreted, and where each backend’s contract lives.

Code generation & export

tvbo export, the available targets, and what the emitted file contains.

Reproducible workflows

tvbo workflow turns a study into a Snakemake DAG whose steps are its experiments, analyses and figures.

The study directory

The BIDS-shaped layout that holds the spec, the code, the results and the report.

The generated code has to run somewhere: ④ Run.

Next