seraplot 2.6.0

Rust data visualization framework - The modern Plotly alternative

SeraPlot - High-Performance Visualization & Production-Ready ML

SeraPlot is a Rust-native framework that bundles 60+ interactive chart types and a complete machine-learning toolkit in a single binary - no Plotly, no Matplotlib, no scikit-learn dependency. Build dashboards, train models, ship to production, all from one library.

📖 Full documentation: https://feur25.github.io/seraplot/introduction.html

📦 PyPI: pip install seraplot ·

---

Why SeraPlot

• 100x to 8000x faster than Plotly + Matplotlib on chart generation - measured, reproducible.
• Zero Python dependencies for plots - every renderer ships as compiled Rust, ready for serverless, edge devices and embedded systems.
• Plots + ML in one package - no need to glue 5 libraries together. Visualize, preprocess, train, evaluate from a single import seraplot as sp.
• Production-grade - online learning (Welford partial_fit), incremental encoders, deterministic outputs, zero hidden global state.
• Multi-target - Python (PyPI), JavaScript / WebAssembly (npm), C / C++ FFI, C# bindings.
• Self-contained HTML output - every chart is a single inlined HTML file. Drop into a slide, an email, a notebook, a dashboard - it just works.

---

Gallery - Chart Types

2D Charts

3D Charts

---

Machine Learning - Production Toolkit

A scikit-learn compatible API, written in Rust, with online-learning primitives ready for streaming workloads.

Supervised Learning

• Linear models: Linear / Ridge / Lasso / ElasticNet / Logistic regression with score, predict_proba, decision_function.
• Trees & ensembles: Decision Tree (classifier + regressor), Random Forest, Gradient Boosting.
• Neighbors: KNN classifier + regressor.
• Naive Bayes, SVM, MLP - full estimator interface.

Unsupervised Learning

• Clustering: KMeans, DBSCAN, Agglomerative, Mean-Shift.
• Dimensionality reduction: PCA, t-SNE, UMAP.

Preprocessing & Pipelines

• Online scalers: StandardScaler with Welford partial_fit for streaming data - no full pass required.
• Incremental encoders: OneHotEncoder and OrdinalEncoder with category union across batches.
• Pipelines: Chain transformers + estimator with full score / predict_proba / decision_function propagation.
• Train/test split, cross-validation, grid search.

Metrics

Accuracy, precision, recall, F1, ROC-AUC, log-loss, R-squared, MAE, MSE, RMSE, confusion matrix, classification report - identical signatures to scikit-learn.

---

Installation

pip install seraplot

Conda / uv users:

conda install -c conda-forge seraplot
uv pip install seraplot

JavaScript / WebAssembly:

npm install seraplot

---

Performance

SeraPlot consistently delivers 100x to 8000x speedups over Plotly and Matplotlib on chart generation, with a flat memory profile that fits inside containers, lambdas and embedded targets. ML estimators are within ~5% of scikit-learn while supporting partial_fit for streaming workloads that scikit-learn cannot handle natively.

---

• Documentation: https://feur25.github.io/seraplot/introduction.html