axiombraid 2.0.0

Explainable, safety-first data quality, validation, cleaning, and drift analysis.

AxiomBraid 2.0

Explainable, safety-first data quality for Python.

AxiomBraid inspects, validates, cleans, compares, and monitors tabular datasets while keeping automated changes conservative, visible, and reproducible.

Installation

pip install axiombraid

Windows:

py -m pip install axiombraid

Optional charts:

pip install "axiombraid[charts]"

Quick start

import axiombraid as AB

result = AB.inspect("students.csv")
AB.report("students.csv")
cleaned = AB.clean("students.csv", risk="low")
AB.export_html("students.csv", "reports/students.html", theme="dark")

Version 2 explainability

result = AB.inspect(
    "students.csv",
    include_confidence=True,
    include_quality_profile=True,
)

AB.report(
    "students.csv",
    include_confidence=True,
    include_quality_profile=True,
)

The quality profile explains Completeness, Uniqueness, Validity, Consistency, and Integrity. Confidence represents evidence strength, not calibrated probability.

Controlled evaluation

clean = AB.read_csv("clean.csv")
corrupted, truth = AB.inject_issues(clean, missing_rate=0.05, duplicate_rate=0.05, invalid_range_rate=0.05, random_state=42)
results = AB.run_evaluation(clean, corruption_config={"missing_rate": 0.05, "duplicate_rate": 0.05, "invalid_range_rate": 0.05, "random_state": 42})

Evaluation metrics operate at issue/column granularity.

Main capabilities

• Missing values, duplicates, constants, identifiers, outliers, suspicious ranges, text inconsistencies, and date-like text
• Explainable dataset, column, and five-dimensional quality scoring
• Evidence-aware confidence with readable console and HTML reports
• Preview-first, risk-classified, reversible cleaning
• Validation contracts, fingerprints, leakage screening, schema comparison, and drift history
• Controlled synthetic corruption with exact ground truth
• Precision, recall, F1, confidence diagnostics, quality-response evaluation, and benchmarks
• Streaming, caching, plugins, batch processing, CLI, and Roman Urdu reports

Safety

AxiomBraid does not silently delete outliers, clip invalid values, drop identifiers, or mutate input DataFrames through the functional API.

Diagnostics

print(AB.about())
print(AB.self_check())
print(AB.compatibility_check())

CLI

py -m axiombraid --version
py -m axiombraid inspect data.csv --confidence --quality-profile
py -m axiombraid evaluate clean.csv --output reports/evaluation
py -m axiombraid benchmark data.csv --repeats 3 --output reports/benchmark.json

License: MIT
Repository: https://github.com/Adil307/AxiomBraid

Source-code documentation with Doxygen

AxiomBraid 2.0.0 includes final Doxygen configuration for modules, classes, functions, docstrings, source browsing, examples, and Graphviz diagrams.

Install the tools on Windows:

winget install --id DimitriVanHeesch.Doxygen -e
winget install --id Graphviz.Graphviz -e

Generate and open the documentation without PowerShell execution-policy issues:

.\generate_doxygen.cmd

The generated main page is:

docs/doxygen/html/index.html

See docs/DOXYGEN.md for setup, troubleshooting, CI artifact generation, and reviewer-sharing instructions.