cli-reference.md

CLI Reference

Binary: datasynth-data

generate

Generate synthetic accounting data.

datasynth-data generate [OPTIONS]

Flag	Description
-c, --config <PATH>	Path to YAML configuration file
-o, --output <DIR>	Output directory (default: ./output)
--demo	Use built-in demo preset
--preset <NAME>	Apply overlay preset (audit-group)
--scenario-pack <NAME>	Load a scenario pack YAML
--fingerprint <PATH>	Generate from a .dsf fingerprint file
--scale <FLOAT>	Scale factor for fingerprint generation (default: 1.0)
-s, --seed <INT>	Random seed for reproducibility
--banking	Enable KYC/AML data generation
--audit	Enable audit engagement generation
--memory-limit <MB>	Memory ceiling in MB (default: 1024)
--max-threads <N>	Max CPU threads (default: half of cores)
--graph-export	Enable PyTorch Geometric graph export
--stream-target <URL>	Stream hypergraph JSONL to endpoint
--stream-api-key <KEY>	API key for stream endpoint
--stream-batch-size <N>	Lines per HTTP POST (default: 1000)
--quality-gate <LEVEL>	Quality gate: none, lenient, default, strict
--fiscal-year-months <N>	Months per fiscal year for multi-period
--append	Append to existing output (requires session.dss)
--months <N>	Additional months for incremental generation
--fraud-scenario <NAME>	Apply fraud scenario (repeatable)
--fraud-rate <FLOAT>	Override fraud rate
--stream-file <PATH>	Stream output to JSONL file
--export-format <FMT>	Additional exports: sap, fec, gobd (repeatable)
--auto-tune	AI-powered generate/evaluate/patch loop
--max-iterations <N>	Max auto-tune iterations (default: 3)

validate

Check a configuration file for errors.

datasynth-data validate --config config.yaml

init

Generate a sample configuration file.

datasynth-data init [OPTIONS]

Flag	Description
-o, --output <PATH>	Output path (default: datasynth_config.yaml)
-i, --industry <NAME>	Industry preset (default: manufacturing)
-c, --complexity <LEVEL>	small, medium, large (default: medium)
--from-description <TEXT>	Generate config from natural language (requires llm feature)

info

Display available presets, industries, and complexity levels.

verify

Verify output integrity (checksums, record counts).

datasynth-data verify --output ./output --checksums --record-counts

fingerprint

Privacy-preserving fingerprint extraction and synthesis. See Fingerprinting.

Subcommand	Description
extract	Extract fingerprint from CSV data to .dsf
validate	Validate a .dsf file
info	Show fingerprint statistics
diff	Compare two fingerprints
evaluate	Evaluate synthetic data fidelity against fingerprint
synthesize	Generate synthetic data from fingerprint

scenario

Counterfactual scenario management. See Scenario Engine.

Subcommand	Description
list	List scenarios in a config
validate	Validate scenario definitions
generate	Generate paired baseline/counterfactual datasets
diff	Compare baseline vs. counterfactual outputs
export	Export scenario as portable .dss file
import	Import .dss scenario into a config

adversarial

ONNX model boundary probing. Requires adversarial feature. See Adversarial Testing.

datasynth-data adversarial --model model.onnx --features 10 --probes 1000

audit

Audit FSM blueprint commands. See Audit FSM Engine.

Subcommand	Description
validate	Validate a blueprint YAML
info	Display blueprint information
run	Run standalone FSM engagement
diff	Compare two blueprints structurally
benchmark	Generate benchmark audit event log