Show the setup: LSD source root and tag, models folder, whether a C++ compiler was found, whether LSD's command-line utilities are built yet (they are built automatically on first use), and whether Rscript and the R package LSDsensitivity are available.

List models. group='models' is the user's LSD_MODELS folder; group='examples' is the Example folder of the LSD distribution (read-only). A model is a folder with an equation file (fun_*.cpp) and at least one .lsd configuration. Returns relative path, title, equation file, source files (*.cpp, *.h, *.hpp in the folder) and configuration names for each.

Return the text of the model's equation file (the C++ source with the EQUATION(...) blocks). Models whose equations are spread over several files list them as source_files in list_models; pass such a plain file name (.cpp, .h or .hpp, no path) as file to read it instead.

Describe a .lsd configuration: the object tree with instance counts, and for each element (variable, parameter or function) its type, number of lags, whether it is saved to the result files, and its value (a single value if all instances are equal, otherwise min, max and count), plus the run settings SIM_NUM, SEED, MAX_STEP and EQUATION. config is the file name with or without .lsd. Set only when they apply: computed=false on objects (not computed), and debug, plot, parallel, saved_separately on elements. For big models pass object='Name' to describe one object, or detail='names' for the tree with instance counts and just the element names grouped by type.

Copy a model folder (source files only: no binaries, results, backups, numbered design configurations, design tables or _sa folders) into the models folder as name, so it can be edited. Fails if name exists.

Replace the model's equation file with content (complete C++ source), or the source file named by file (plain name, .cpp, .h or .hpp; a new file is created if it does not exist). The version before the latest write is kept as .bak and the one before the first write as .orig. Only models in the models folder can be written.

Set element values in a configuration, using LSD's own lsd_confgen. values maps element name to a number; every instance of the element gets the value. A variable's name sets its initial value at the first lag; "Name -2", "Name -3" ... (name, space, negative lag) set the second, third ... lag ("Name -1" is the first); a lag beyond the variable's lags, or on a parameter, is an error, and one element can appear once per call. Writes new_config.lsd, or replaces config.lsd (the old file is kept as .bak). The result says whether an existing file was replaced (replaced, backup).

Edit SIM_NUM (runs), SEED and MAX_STEP (time steps) of a configuration.

Mark the named variables or parameters as saved (or not saved) to the result files. Only saved elements appear in results.

Change a configuration's structure (objects, parameters, variables, functions, instance counts) with LSD's own code. operations is a list of objects applied in order; if any fails (the message names it) nothing is written. Writes new_config.lsd, or replaces config.lsd (old file kept as .bak). Names must be valid LSD labels (letter or '' first, then letters, digits, '') and unique across the whole model, objects and elements alike. Operations, one example each: {"op": "add_object", "parent": "Root", "name": "Firm", "instances": 10} {"op": "add_parameter", "object": "Firm", "name": "alpha", "value": 0.5} {"op": "add_variable", "object": "Firm", "name": "K", "lags": 1, "initial": 1.0, "saved": true} (initial: a number, or one per lag) {"op": "add_function", "object": "Firm", "name": "f"} {"op": "rename", "name": "old", "new_name": "new"} (object or element) {"op": "delete", "name": "K"} (element; an object with elements or child objects needs "force": true and goes with all of it) {"op": "set_instances", "object": "Firm", "instances": 50} {"op": "set_instance_values", "name": "alpha", "values": [0.1, 0.2], "lag": 2} (one value per instance in file order; lag only for variables) {"op": "describe", "name": "alpha", "text": "adjustment speed"} Parameters and variables are added to every instance of the object, with the value (default 0) in each. set_instances gives the object that number under every parent instance; new instances are copies of the object's first instance (its values and its child objects), extra ones are removed from the end; at least 1. Equations are not touched: add or change them with write_equations. LSD's save rewrites the file in its own layout, so old files may change in layout (empty descriptions, the notes LSD generates for initial values of variables with no lags are dropped), never in values.

Create a new, empty model folder in the models folder, as LSD's model manager (LMM) does: equation file fun_.cpp from LSD's template, model_options.txt, modelinfo.txt (LMM lists a folder only with it), description.txt, and a configuration Sim1.lsd holding only Root. name is the folder (letters, digits, underscores; it must not exist or lie inside another model). Then use edit_structure to add objects, parameters and variables, write_equations for the equations, set_saved if needed, and run_configuration. An empty model compiles and runs (it computes nothing and saves nothing).

Compile the model's equation file into a headless LSD program (built into the cache, not the model folder). Returns ok and build time, or the first compiler errors as file:line: message. Nothing is rebuilt if the equation file is unchanged.

Compile if needed and run a configuration of a model in the models folder. seed and runs override SEED and SIM_NUM of the file. threads: with runs > 1, the number of runs executed in parallel; otherwise threads for models that use parallel objects. Results are written next to the configuration as .res.gz. Returns the files written and, for the first run, the last value and mean of each saved series (at most 50 series, those whose name has few instances first; the rest are counted). With runs > 1 (sequential) LSD also writes _.tot.gz; with threads set the runs are parallel and no totals file is written. Row 0 of a result file holds initial values. On failure returns the tail of LSD's output.

Read time series from one result file (.res.gz) in the model folder. variables filters by name ('Mean') or name with instance ('Mean 1'); start/end select time steps (the row number in the file is the time step; row 0 holds initial values; start must not exceed end; max_points >= 1); the series are thinned evenly to max_points. At most 50 series are returned.

Create a design of experiments for a sensitivity analysis and write it in LSD's own file layout in the model folder: .sa, the design table _1_N.csv (and for meta-model designs the out-of-sample table _N+1_N+V.csv), numbered configurations _1.lsd ..., and _design.json (our file: method and parameters). factors maps an element to [min, max], or [min, max, "int"] for integers. An element is a parameter name, or a variable name for the variable's initial value at its first lag, or "Name -2", "Name -3" ... (name, space, negative lag) for the initial value at that lag. Functions, variables with no lags and lags beyond the variable's lags are refused, and an element can be a factor only once (LSD's design table names a factor by its label, so the result files and the analysis tables show the plain name, for example "X"; the design file _design.json records each factor's lag, 0 for a parameter). method: 'lhs' (Latin hypercube) or 'random': samples points (required, at least 2) plus validation_samples uniform out-of-sample points, for a Kriging or polynomial meta-model. 'nolh': near-orthogonal Latin hypercube made by LSD's own NOLH tables. LSD chooses the table from the number of factors (17 points for 1-7 factors, 33 for 8-11, 65 for 12-16, 129 for 17-22, 257 for 23-29, 512 for 30-100; extended=True uses LSD's extended size for the table: 33, 65, 129, 257, 257, 512), so samples is not used; the result says how many points LSD produced. Plus validation_samples out-of-sample points by LSD's Monte Carlo range sampling. For meta-model analysis. 'ee': elementary effects (Morris) design made by LSD's own code: trajectories (default 10) each of factors + 1 points, chosen from a pool of pool random trajectories (default 100) on levels levels (even, default 4) with jump (default 2). No out-of-sample set; validation_samples and samples are ignored. An ee design made on macOS differs from one made on Linux for the same seed (the C++ library's shuffle differs); both are valid designs. Analyse it with sa_analyze (metamodel 'ee'). Each point runs runs_per_point times (at least 2) with its own seeds. seed seeds the sampling. Refuses to replace an existing design unless overwrite=True. A warning is returned for integer factors with few levels (a hypercube collapses onto them). Variables whose results will be analysed must be saved (set_saved).

Run every numbered configuration of the design in parallel processes (default: one per CPU). Points whose result files already exist are skipped. Returns counts of points done, run, failed and not started.

Analyse the design results for one saved variable with LSD's R package LSDsensitivity. metamodel 'kriging' (default for lhs, random and nolh designs) or 'polynomial' fits a meta-model and computes its Sobol decomposition: returns the fit quality (Q2 for kriging, R2 for polynomial) and a table with direct effects and interactions per factor. For a design made with method 'ee' the analysis is elementary effects (metamodel 'ee', chosen automatically): returns per factor mu, mu_star, sigma, se and p_value (parameters scaled to [0, 1]; mu_star is the overall effect, sigma non-linear or interaction effects, p_value tests mu_star = 0), sorted by mu_star. Kriging or polynomial on an ee design, or ee on another design, is an error. For an ee design made in LSD's own interface (no design file) pass metamodel='ee' with its levels and jump. ini_drop drops initial time steps, n_keep keeps that many (-1 = all). The response is the mean of the variable over the kept steps, averaged over runs. Variables whose name starts with '_' work; with several instances only the first instance is analysed. ini_drop must be below MAX_STEP and ini_drop + n_keep at most MAX_STEP. r_seed seeds R's random numbers, so identical calls give identical results. A warning is added when a meta-model fit is below 0.5. The polynomial meta-model fails when a design point has a negative mean response (LSD weights points by mean/SD) and needs at least two factors; use kriging then. Kriging can fail numerically when design points nearly coincide (the message says so; try polynomial). Needs Rscript and LSDsensitivity; says so if they are missing.