Featuresmith v0.4.0 — Recommendation & Planning

What Featuresmith can do today (v0.4.0)

• Recommendation Engine: Merges every review finding into a single ranked, explainable list with deterministic confidence scores and full traceability back to originating findings and reviewers
• Plan Primitive: fs.plan() and featuresmith plan compile accepted recommendations into an inspectable, deterministic Plan of transformation steps
• FeatureQualityReviewer: 10th built-in reviewer detecting near-constant, redundant, and low-signal columns
• Dataset Review Engine: 10 automated reviewers evaluating schema, missingness, duplicates, distributions, feature quality, target leakage, and snapshot deltas
• ML Readiness Score: Explainable 0–100 composite scorecard across 7 effective health dimensions with fix suggestions (Class Balance omitted until its detector ships)
• Score Dimension Reconciliation: Consistency and Data Quality dimensions consolidated, eliminating the cardinality double-count
• Intelligent Leakage Detection: 6 pattern detectors (target correlation, identifier shape, timestamp anomalies, duplicate targets)
• Dataset Diff Engine: Snapshot comparisons via fs.diff() and featuresmith diff command for schema, nulls, and health deltas
• Supported Data Sources: CSV, Parquet, Excel, pandas, and Polars DataFrames
• Python SDK: Public entry points for load(), profile(), analyze(), review(), diff(), score(), and plan()
• CLI Interface: featuresmith analyze, review, diff, and plan commands with Rich table, JSON output, and exit-code CI gating