featuresmith-core

featuresmith-core is the Python SDK and deterministic analysis engine for Featuresmith — the open-source Dataset Review Platform for structured data.

It provides automated dataset code reviews, 0–100 ML readiness scores, target leakage detection, version snapshot diffing, and deterministic transformation planning.

Key Capabilities

• Dataset Ingestion (fs.load()): Native support for CSV, Excel, Parquet, pandas, and Polars DataFrames.
• Automated Dataset Review (fs.review()): 10 built-in reviewers evaluating schema health, missingness, duplicates, constants, cardinality, statistics, target leakage, snapshot deltas, and feature quality.
• ML Readiness Score (fs.score()): An explainable 0–100 quality scorecard calculated across 7 effective health dimensions.
• Intelligent Leakage Detection: 6 named pattern detectors recognizing target correlation, identifier shape, timestamp anomalies, and duplicate targets.
• Dataset Diff Engine (fs.diff()): Version snapshot comparison engine surfacing schema drift, null spikes, and quality regressions between two datasets.
• Recommendation Engine & Plan Primitive (fs.plan()): Centralized engine merging review findings into ranked fix recommendations and compiling accepted items into inspectable, deterministic Plan objects.

Installation

pip install featuresmith-core

Quick Start (Python SDK)

import featuresmith as fs

# 1. Load dataset
dataset = fs.load("data/train.csv")

# 2. Run automated dataset code review
review_res = fs.review(dataset, target_column="target")

# 3. Extract 0-100 ML Readiness Scorecard
scorecard = fs.score(review_res)
if scorecard:
    print(f"ML Readiness Score: {scorecard.overall}/100")

# 4. Compile an inspectable Plan from accepted recommendations
plan = fs.plan(review_res, accept=["rec.quality.missingness.cabin"])
for item in plan.items:
    print(f"Plan Step: {item.title} (confidence {item.confidence})")

For comprehensive guides and API reference, visit the official website: https://featuresmith.adityagangwani.me