ClarivueXAI: Unified Explainable AI Framework

ClarivueXAI is a Python library for explainable AI, providing a unified API to explain models across frameworks (scikit-learn, TensorFlow, PyTorch, custom) and data types (tabular, text, images, time series). Problem Explainable AI is fragmented, with different tools for different models and data types, leading to inconsistent APIs and slower workflows for data scientists and ML engineers. Solution ClarivueXAI unifies XAI with:

A consistent API for all model types and explanation methods. Support for multiple frameworks and data types. Global and local explanations using methods like SHAP, LIME, and Integrated Gradients. Advanced visualization tools, including interactive plots.

See Examples for detailed demonstrations. Key Features

Unified API: Consistent interface for all models. Multi-modal Support: Handles tabular, text, image, and time series data. Framework Agnostic: Supports scikit-learn, TensorFlow, PyTorch, and custom models. Global & Local Explanations: Model-level and instance-level insights. Advanced Visualization: Static and interactive plots, as shown in Wine Classification.

Installation

Base installation

pip install clarivuexai

TensorFlow support

pip install clarivuexai[tensorflow]

PyTorch support

pip install clarivuexai[torch]

All dependencies

pip install clarivuexai[all]

Quick Example This example, inspired by Random Forest Example, shows how to explain a scikit-learn model: from clarivuexai import Model, Explainer from sklearn.ensemble import RandomForestClassifier from sklearn.datasets import load_iris import matplotlib.pyplot as plt

Load data

iris = load_iris() X, y = iris.data, iris.target

Train model

model = RandomForestClassifier(n_estimators=100, random_state=42) model.fit(X, y)

Wrap with ClarivueXAI

cxai_model = Model.from_sklearn(model, feature_names=iris.feature_names)

Create explainer

explainer = Explainer(cxai_model)

Get explanations

global_exp = explainer.explain_global(X, method='shap') local_exp = explainer.explain_local(X[0:1], method='shap')

Visualize

plt.figure(figsize=(10, 6)) global_exp.plot() plt.title("Global SHAP Feature Importance") plt.tight_layout() plt.savefig("global_shap.png")

Documentation Full documentation is available at clarivuexai.readthedocs.io. Key sections:

Quickstart Guide: Basic setup. API Reference: Detailed API docs. Examples: Use cases for tabular, text, image, and time series data.

Contributing Contributions are welcome! Submit a Pull Request on GitHub. License Licensed under the MIT License. See the LICENSE file.