Python SDK
Data Models
SDK Reference: the complete typed output schema
Every Featuresmith result is a Python dataclass with frozen=True and slots=True, so instances are read-only, fast, and safely serializable. Each top-level result object exposes a to_dict() method that produces a plain, JSON-ready dictionary.
The models are grouped by the engine that produces them. Use the pages below for the full field-by-field reference:
Profile ModelsDatasetSummary, ColumnProfile, the four typed column profiles, the three aggregates, and both metadata records returned by fs.profile().Rule & Finding ModelsRuleResult, RuleFinding, finding severities, and the eight built-in validation rules with their default thresholds.Review ModelsReviewResult, ReviewSection, the six ReviewCategory values, the four Severity levels, and the ten built-in reviewers.Score ModelsMLReadinessScore, DimensionScore, the seven effective scoring dimensions, and the deduction formula behind the 0-100 scorecard.Leakage ModelsLeakageFinding and the six pattern detectors that flag target leakage.Diff ModelsDatasetDiffResult and every nested delta model produced by fs.diff().
Related References
- The normalized input model is documented on the Dataset page.
- Error classes raised during ingestion are documented on the Exceptions page.