Commands

Every fobis command, by area. Each page lists all the options of the command (generated from the command line itself); fobis <command> --help prints the same. The fobos file holds the settings a command reads from the project, the user configuration those of the machine.

Building

Command	What it does
build	Build all programs found or specific target(s).
clean	Clean project: remove all OBJs and MODs files. Use carefully.
run	Build a target (if needed) and execute it.
install	Install previously built files, or clone+build+install a GitHub-hosted FoBiS project.
rule	Execute special rules or user-defined ones from a fobos file.
cache	Manage the FoBiS build artifact cache.
introspect	Emit machine-readable JSON project metadata.
tree	Print the resolved inter-project dependency tree.
check	Validate the dependency graph without building.

Dependencies and projects

Command	What it does
new	Create a project: a module in src/, a program in app/, a test in test/, a fobos with debug and release modes and a [test] section, a .gitignore and a git repository (nothing committed). An existing file is never overwritten.
fetch	Fetch and build GitHub-hosted Fortran dependencies listed in the fobos [dependencies] section.
fpm	Interoperate with the Fortran Package Manager: fpm.toml to fobos and back.
scaffold	Manage project boilerplate via bundled scaffold templates.
ecosystem	Manage a set of interconnected FoBiS projects (registry: [ecosystem] in the user config).

Testing

Command	What it does
test	Discover, build, and run Fortran test programs.
doctests	Test all valid doctests snippets found.
coverage	Generate HTML/XML coverage reports from instrumented test runs.

Verifying results

Command	What it does
case	Run the cases declared in the fobos ([case:NAME]) and extract their quantities.
provenance	Print what a result produced now would depend on: commit, dependency pins, compiler, flags, machine.
verify	Run cases and compare their quantities with the stored references, within each quantity's tolerance.
matrix	Build, test (and verify) every mode x shadow combination available on this machine.
assess	Assess the numerical error and uncertainty of a case's quantities from ensembles of its variants.
diff-run	Run the cases on a revision and on the working tree, in scratch copies, and compare their quantities.
bench	Measure the run time of cases: warm-up, then runs until the median's 95% interval is narrow enough; compared with the case's recorded history or, with --compare, with a revision. Faster or slower only when the intervals do not overlap.
gate	One verdict for a change: build, review of the changed lines, tests, verify, diff-run and bench against the base, assessment budgets and the declared matrix, each reduced to pass/fail/skipped with its first failing fact. Exit 1 when a non-advisory stage fails.
profile	Profile a case: where it spends its time (gprof), its instructions (callgrind) or its GPU time and transfers (nsys), mapped to the project's procedures and source lines.

Code analysis

Command	What it does
lint	Report bug-prone and obsolescent Fortran constructs (rules: fobis lint --list-rules).
review	Report the lint findings on the changed lines only (staged, since a base, or uncommitted).
query	Answer a question about the code from the analysis: who calls what, who writes a variable, where a name is defined, which directive regions hold a line, where a variable lives on the device, which MPI calls are made.
modernize	Apply the automatic lint fixes (array constructors, relational operators, generic intrinsics, character length, labelled DO loops) to a scratch copy and keep them only when it builds, the tests still pass and the declared cases give equivalent results; each rule is checked alone when the whole set fails.
bindings	Generate bind(c) shims (NAME_c.f90), a C header (NAME.h) and a Python ctypes wrapper (NAME.py) for the public procedures of the modules whose arguments allow it; report why the others are not bound.
agents-md	Generate the build, test and verification instructions for coding agents (AGENTS.md) from the fobos and the sources, between fobis:begin/end markers: the rest of the file is kept.

AI assistance

Command	What it does
llm	Manage the local LLM: models pulled from the Hugging Face Hub (nothing is downloaded implicitly).
ask	Answer a question from the knowledge packs (and, with --project, the project's procedures), citing sources.
kb	Search the knowledge packs (Fortran, OpenMP, OpenACC, MPI references; your own packs).
testgen	Generate a test program for a module's public procedures; kept only when it builds, runs and checks.
merge-assist	Propose resolutions for merge, rebase or cherry-pick conflicts, checked by the compiler; never stages.
commit	Generate a Conventional Commits message for staged changes via a local LLM.
mcp	Serve lint, rules, knowledge packs and build commands to MCP clients over stdio.