semsynth

Toolkit to profile, describe and synthesize tabular datasets.

Unified model interface: run both Metasyn, PyBNesian and SynthCity models from a single config.
Uniform outputs: each model writes artifacts under dataset/models/<model-name>/.
Provider-aware metadata and UMAP visuals.
Generate static HTML reports

Functions

Return the installed SemSynth version string.

semsynth.get_version() → str: Return the installed SemSynth version string.

Modules

`app`	Minimal Flask application exposing SemSynth search and report actions.
`backends`	Backend implementations and shared interfaces.
`catalog`	Build DCAT catalogs for SemSynth reports and datasets.
`datasets`
`downstream_fidelity`	Downstream fidelity comparison between real and synthetic data.
`jsonld_to_rdfa`	jsonld_to_rdfa.py Generate static HTML with RDFa from nested compact JSON-LD.
`mappings`
`metadata`
`metrics`
`missingness`	Missingness modeling utilities for backend generators.
`models`	Model configuration loading and run discovery utilities.
`pipeline`	Pipeline orchestration for dataset processing and reporting.
`privacy_metrics`
`reporting`	Utilities for rendering dataset reports.
`reports_index`	Stub module to satisfy autosummary; full report index generation lives elsewhere.
`runtime`	Runtime helpers for resolving SemSynth dependencies.
`semmap`
`specs`	Shared data structures for SemSynth dataset handling.
`templates`	Packaged report templates.
`torch_compat`	Compatibility helpers for optional PyTorch features.
`umap_utils`	Utilities for building and rendering UMAP embeddings lazily.
`utils`