data-fitcheck 2.0.1

Zero-boilerplate ML data validation and model evaluation.

FitCheck

Zero-boilerplate ML data validation and model evaluation.

---

FitCheck validates datasets, evaluates ML models, and detects distribution drift — all with one-line commands that generate shareable HTML reports.

Package	Release	Stats
data-fitcheck	pip install data-fitcheck

Philosophy

• Zero Config — Pass a file path. Get answers. No YAML, no setup.
• Immutability — FitCheck diagnoses, never silently modifies. Fix scripts are transparent and inspectable.
• Shareability — Every check generates a self-contained HTML report ready for Slack, email, or GitHub.
• No Telemetry — Zero outbound network calls. All computation is local.

Installation

pip install data-fitcheck

From source (for development):

git clone https://github.com/neoline361-art/fitcheck.git
cd fitcheck
pip install -e ".[dev]"

Quick Start — 30 Seconds

import fitcheck

# 1. Validate a dataset
issues = fitcheck.check("data.csv", target="label")
# → Opens fitcheck_report.html with issues, stats, preview

# 2. Evaluate a model
metrics = fitcheck.report(model, X_test, y_test)
# → Opens model_report.html with metrics + plots

# 3. Detect drift
results = fitcheck.detect_drift("train.csv", "production.csv")
# → Opens drift_report.html with per-feature drift results

CLI (same thing, no Python needed)

pip install data-fitcheck
fitcheck check data.csv --target label
fitcheck drift train.csv production.csv
fitcheck demo  # runs everything in one command

What FitCheck Checks

Check	Method	Severity
Missing values	Null ratio >5% / >20%	Warning / Critical
Duplicate rows	df.duplicated().sum()	Warning
Constant columns	Single unique value	Warning
Class imbalance	Majority class >80%	Warning
Outliers	IQR method (1.5×)	Info
Numeric drift	KS test (p<0.05)	Critical
Categorical drift	Chi-squared (p<0.05)	Critical

Documentation

Resource	Description
API Reference	Complete API documentation with parameters, returns, and examples
Architecture	Module design and design decisions
Design Decisions	Why certain technical choices were made
FAQ	Frequently asked questions
Examples	Runnable usage examples
Benchmarks	Performance benchmarks on standard hardware

Project Maturity

Aspect	Status
Version	v2.0.0 — Semantic Versioning
Tests	30 tests, 82% coverage
Type Safety	mypy strict — 11 type-arg warnings (all ndarray, no runtime impact)
Linting	ruff clean
Security	bandit + pip-audit (pickle warning expected for model loading)
License	Apache 2.0
Platforms	Linux, macOS, Windows (Python 3.9–3.13)
PyPI	v2.0.0 — pip install data-fitcheck

Limitations

• Supports only pandas DataFrames (CSV, Parquet)
• Drift: KS test for numeric, Chi-squared for categorical
• Deep learning model evaluation is not implemented
• Datasets must fit in memory
• No streaming or distributed processing

Development

# Setup
pip install -e ".[dev]"
pre-commit install

# Run all quality gates
ruff check fitcheck/
mypy fitcheck/
bandit -r fitcheck/ -x tests
pip-audit
pytest --cov=fitcheck --cov-report=term-missing

# Demo
python demo.py

Versioning

FitCheck follows Semantic Versioning. Given a version MAJOR.MINOR.PATCH:

• MAJOR — Breaking API changes
• MINOR — New features, backward compatible
• PATCH — Bug fixes, backward compatible

See CHANGELOG.md for per-release details.

Contributing

See CONTRIBUTING.md. TL;DR:

1. Tests must pass
2. Ruff + mypy must be clean
3. Add CHANGELOG entry

Security

See SECURITY.md. Report vulnerabilities to [e-mail].

License

Apache 2.0 — see LICENSE for details.