🧙‍♂️ SciWizard

---

Features

Area	What you get
Data	CSV loading, table preview, data profiling, missing value handling
Preprocessing	Label/one-hot encoding, column dropping, scaler info
Visualisation	Histograms, scatter plots, correlation heatmaps, feature distributions, PCA 2D
Training	7 classification + 7 regression algorithms, hyperparameter control, CV scores
AutoML	Automatic model sweep with sortable leaderboard
Evaluation	Confusion matrix, ROC curve, cross-validation bar chart, metrics dashboard
Prediction	Form-based single prediction, batch CSV prediction with export
Registry	Persistent model save/load/delete with metadata
Experiments	JSONL-backed run history with full metric tracking
Plugins	Drop Python files into /plugins to add custom models and preprocessors

---

Installation

Requirements: Python 3.10+

# Clone
git clone https://github.com/pro-grammer-SD/sciwizard.git
cd sciwizard

# Create virtual environment
python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate

# Install
pip install -e ".[dev]"

---

Usage

python -m sciwizard

Or use the installed entry point:

sciwizard

Basic workflow

1. Data tab — load a CSV, select target column, handle missing values
2. Preprocess tab — encode categoricals, drop irrelevant columns
3. Visualize tab — explore distributions and correlations
4. Train tab — pick algorithm, configure split, train
5. Evaluate tab — inspect confusion matrix, ROC, CV scores
6. Predict tab — enter values for single prediction or upload a batch CSV
7. Registry tab — save and reload trained models
8. Experiments tab — review all past runs

---

Writing a Plugin

Drop a .py file into the plugins/ directory:

# plugins/extra_trees.py
from sklearn.ensemble import ExtraTreesClassifier

def register(registry: dict) -> None:
    registry["models"]["Extra Trees"] = ExtraTreesClassifier(n_estimators=100)

SciWizard discovers it on next launch and adds it to the model selector.

---

Running Tests

pytest

---

Project Structure

sciwizard/
├── sciwizard/
│   ├── app.py                  # Bootstrap & main()
│   ├── config.py               # Constants & paths
│   ├── core/
│   │   ├── data_manager.py     # CSV loading, profiling, cleaning
│   │   ├── model_trainer.py    # Training, evaluation, AutoML
│   │   ├── model_registry.py   # Persistent model storage
│   │   ├── experiment_tracker.py
│   │   └── plugin_loader.py
│   └── ui/
│       ├── main_window.py      # Top-level window + sidebar
│       ├── theme.py            # Dark stylesheet
│       ├── workers.py          # QThread/QRunnable wrappers
│       ├── panels/             # One file per application tab
│       └── widgets/            # Shared UI components
├── tests/
├── docs/
├── plugins/                    # Drop custom model plugins here
└── icon/
    └── icon.ico

---

Contributing

See CONTRIBUTING.md.

License

MIT — see LICENSE.