Metadata-Version: 2.4
Name: optiflowx
Version: 0.0.7
Summary: A modular optimization and machine learning workflow framework for hyperparameter tuning and performance benchmarking across diverse datasets.
Project-URL: Homepage, https://github.com/Faycal214/optiflowx
Project-URL: Repository, https://github.com/Faycal214/optiflowx
Project-URL: Issues, https://github.com/Faycal214/optiflowx/issues
Author-email: Alikacem Faycal <faycal213.dz@gmail.com>
License: MIT
License-File: LICENSE
Keywords: AutoML,ant colony optimization,bayesian optimization,combinatorial optimization,genetic algorithm,grey wolf optimization,hyperparameter optimization,machine learning,particle swarm optimization,simulated annealing,tpe
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.10
Requires-Dist: numpy
Requires-Dist: optuna
Requires-Dist: scikit-learn
Requires-Dist: scikit-optimize
Requires-Dist: tqdm
Provides-Extra: dev
Requires-Dist: black>=24.0; extra == 'dev'
Requires-Dist: build; extra == 'dev'
Requires-Dist: mypy>=1.5; extra == 'dev'
Requires-Dist: pytest-cov>=4.0; extra == 'dev'
Requires-Dist: pytest>=7.0; extra == 'dev'
Requires-Dist: ruff>=0.5.0; extra == 'dev'
Requires-Dist: twine; extra == 'dev'
Provides-Extra: docs
Requires-Dist: mkdocs; extra == 'docs'
Requires-Dist: mkdocs-material; extra == 'docs'
Provides-Extra: optuna
Requires-Dist: optuna>=3.0; extra == 'optuna'
Provides-Extra: skopt
Requires-Dist: scikit-optimize>=0.9; extra == 'skopt'
Provides-Extra: xgboost
Requires-Dist: xgboost>=2.0; extra == 'xgboost'
Description-Content-Type: text/markdown

# OptiFlowX

[![PyPI version](https://img.shields.io/pypi/v/optiflowx)](https://pypi.org/project/optiflowx/)
[![Python versions](https://img.shields.io/pypi/pyversions/optiflowx.svg)](https://pypi.org/project/optiflowx/)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![Build Status](https://github.com/Faycal214/optiflowx/actions/workflows/test.yml/badge.svg)](https://github.com/Faycal214/optiflowx/actions)

OptiFlowX is a production-oriented, modular framework for hyperparameter and configuration optimization. It combines metaheuristic and probabilistic search algorithms with flexible search-space abstractions, an easy-to-use wrapper for models, and efficient parallel evaluation—designed for both research and production use.

Current PyPI release: `0.0.7` — https://pypi.org/project/optiflowx/

Key capabilities:

- Optimize models for classification and regression tasks
- Support for scikit-learn models and user-supplied custom models
- Built-in metrics and user-provided custom metric callables
- Built-in search-space definitions and fully custom search spaces
- Parallel candidate evaluation with optional `dill` fallback for non-pickleable callables

Table of contents

- [Why OptiFlowX](#why-optiflowx)
- [Key features](#key-features)
- [Algorithms included](#algorithms-included)
- [Quickstart](#quickstart)
- [Examples](#examples)
- [Search space](#search-space)
- [Parallelism & multiprocessing notes](#parallelism--multiprocessing-notes)
- [API reference (quick)](#api-reference-quick)
- [Development & testing](#development--testing)
- [Contributing](#contributing)
- [License](#license)
- [Contact & citation](#contact--citation)

Why OptiFlowX

- Single, consistent API across many optimizers (metaheuristics and probabilistic methods)
- Production-minded: packaged with docs, examples, tests, and CI
- Extensible — add new optimizers, model wrappers, or search-space primitives

Key features

- Unified optimizer interface (PSO, GA, ACO, GWO, SA, TPE, Bayesian, Random Search)
- Classification and regression support
- Plug-and-play model configs (`optiflowx.models.configs.*`) and `ModelWrapper` for CV and final fitting
- Built-in metric helpers and a `get_metric()` abstraction that normalizes regression metrics for maximization
- Flexible `SearchSpace` supporting continuous, discrete, and categorical parameters with sampling and grid generation
- Parallel evaluation via `ParallelExecutor` with pickle/dill serialization fallbacks

Algorithms included

- Metaheuristics / Combinatorial:
  - Particle Swarm Optimization (PSO)
  - Genetic Algorithm (GA)
  - Simulated Annealing (SA)
  - Ant Colony Optimization (ACO)
  - Grey Wolf Optimizer (GWO)
- Probabilistic / Bayesian & related:
  - Tree-Structured Parzen Estimator (TPE)
  - Bayesian Optimization
  - Random Search

All algorithms are implemented under `optiflowx.optimizers.*`. Most optimizers expose a `run(max_iters=...)` method that returns `(best_params, best_score)`.

Quickstart

Install (recommended inside a virtual environment):

```bash
python -m venv venv
source venv/bin/activate
pip install -e .
# Optional extras
pip install dill xgboost
```

Minimal example (random forest + PSO):

```python
from sklearn.datasets import make_classification
from optiflowx.models.configs.random_forest_config import RandomForestConfig
from optiflowx.optimizers.pso import PSOOptimizer

X, y = make_classification(n_samples=200, n_features=12, random_state=0)
cfg = RandomForestConfig()
wrapper = cfg.get_wrapper(task_type="classification")

opt = PSOOptimizer(
    search_space=cfg.build_search_space(),
    metric="accuracy",
    model_class=wrapper.model_class,
    X=X, y=y,
    n_particles=12,
)
best_params, best_score = opt.run(max_iters=10)
print(best_score, best_params)
```

Examples

The `examples/` directory contains runnable scripts covering common combinations:

- classification with sklearn models and sklearn/custom metrics
- regression with sklearn models and sklearn/custom metrics
- examples that use `CustomModelConfig` for user-defined model wrappers

Run one of the example scripts directly:

```bash
python examples/classification/classification_sklearn_model_sklearn_metric.py
```

Search space

Use built-in configs for quick starts (e.g., `RandomForestConfig().build_search_space()`), or create custom spaces with `optiflowx.core.search_space.SearchSpace`:

```python
from optiflowx.core.search_space import SearchSpace

s = SearchSpace()
s.add("n_estimators", "discrete", [10, 50, 100, 200])
s.add("learning_rate", "continuous", [1e-3, 0.3], log=True)
s.add("criterion", "categorical", ["gini", "entropy"])
```

Parallelism & multiprocessing notes

`ParallelExecutor` uses `multiprocessing.Pool` to evaluate candidates concurrently. If you pass non-pickleable callables (e.g., nested functions or closures) as custom metrics, the executor will:

1. Try to serialize with `pickle`.
2. If `pickle` fails and `dill` is installed, it will serialize using `dill`.
3. If serialization is not possible and multiple workers are requested, the executor raises an error. If only one worker is used it will fall back to sequential evaluation.

If you plan to pass nested custom metrics and want parallel execution, install `dill`:

```bash
pip install dill
```

API reference — quick

High-level building blocks (see docstrings for full signatures):

- `optiflowx.core.search_space.SearchSpace` — define and sample hyperparameter spaces
- `optiflowx.core.model_wrapper.ModelWrapper` — cross-val evaluation and final fitting
- `optiflowx.core.metrics.get_metric` — normalized metric callables (negates regression errors so optimizers maximize)
- `optiflowx.core.parallel_executor.ParallelExecutor` — parallel candidate evaluation
- `optiflowx.optimizers.*` — concrete optimizers (e.g., `PSOOptimizer`, `GeneticOptimizer`)
- `optiflowx.models.configs.*` — model configs exposing `build_search_space()` and `get_wrapper()`

Typical high-level flow:

1. Select model config from `optiflowx.models.registry.MODEL_REGISTRY`.
2. Build its `SearchSpace` and `ModelWrapper`.
3. Initialize an optimizer with the space, metric, `model_class`, and data.
4. Run `optimizer.run(max_iters=...)` or use `optiflowx.core.OptimizationEngine` / `MLPipeline` to orchestrate runs.

Development & testing

Run tests locally (recommended in a virtualenv):

```bash
pip install -r requirements-test.txt
pytest -q
```

Developer tools (optional):

```bash
pip install black ruff mypy pytest pytest-cov
ruff check .
black .
mypy optiflowx
```

Contributing

Contributions welcome. Please follow these steps:

1. Open an issue describing the feature or bug.
2. Create a topic branch in your fork.
3. Add tests for any new behavior.
4. Submit a PR with a clear description and changelog entry.

If you plan to add new optimizers or model configs, aim for:

- clear docstrings and examples under `examples/`;
- unit tests in `tests/` exercising the integration (optimizer + wrapper + executor);
- lightweight, focused commits.

License

This project is licensed under the MIT License — see the `LICENSE` file for details.

Contact & citation

If you use OptiFlowX in research or production, please cite:

```bibtex
@software{optiflowx,
    author = {Faycal, Alikacem},
    title = {OptiFlowX: Combinatorial Hyperparameter Optimization Framework},
    year = {2025},
    url = {https://github.com/Faycal214/optiflowx}
}
```

Contact:

- Author: Alikacem Faycal
- Email: faycal213.dz@gmail.com

Acknowledgements

This project draws on many open-source tools and libraries including scikit-learn, Optuna, scikit-optimize, and others. See `pyproject.toml` for declared dependencies.

----

If you'd like, I can also:

- run the test suite in this environment and report results,
- produce a condensed one-page quick reference for the most common API calls,
- add badges for coverage/Docs/CodeQL if you want to include them in CI.


