Core Concepts

Mental Model & Workflow

How Featuresmith capabilities fit together

Featuresmith is organized into progressive, modular layers. Understanding how these functions relate helps you choose the right API for your data pipeline.

The Main Execution Flow

bash

1[Raw CSV / Parquet / Excel / DataFrame]
2               │
3               ▼  fs.load()
4        [Dataset Object]
5               │
6   ┌───────────┼───────────┐
7   ▼           ▼           ▼
8fs.profile() fs.analyze() fs.review()
9 (Stats)     (Rules)     (Review + Score)
10                           │
11                           ▼  fs.score()
12                    [MLReadinessScore]
13                           │
14                           ▼  fs.diff()
15                  [DatasetDiffResult]

Which Function Should I Use?

"How do I load data into a clean, standard wrapper?"

Use ds = fs.load(source) to parse files or DataFrames into a normalized Dataset.

"How do I extract raw statistical summaries without running quality rules?"

Use prof = fs.profile(ds) for min, max, mean, missingness, cardinality, and correlation summaries.

"How do I check atomic rule assertions?"

Use res = fs.analyze(ds, target_column=...) to get flagged RuleFinding objects.

"How do I run a comprehensive automated dataset code review?"

Use rev = fs.review(ds, target_column=...) to run 10 reviewers and get structured sections.

"How ready does the dataset appear for machine learning?"

Use scorecard = fs.score(rev) to extract an explainable 0–100 quality scorecard.

"How did two dataset snapshot versions change?"

Use diff_res = fs.diff(v1, v2) to compare snapshots and get an overall health verdict.