Explore Documentation Modules
SwiftDataFrame →
Zero-copy columnar tables built on Apache Arrow. Features pure-Swift Parquet reader/writer, out-of-core ChunkedDataFrame, POSIX mmap partitioning, and SIMD hash joins.
SwiftStats →
Vectorized descriptive statistics, probability distributions (Student-t, Chi-Square, F), paired t-tests, and ANOVA powered by Accelerate vDSP.
SwiftPreprocessing →
StandardScaler, MinMaxScaler, OneHotEncoder, TargetEncoder, KNNImputer, and scikit-learn parity Pipeline & ColumnTransformer.
SwiftML →
Linear/Logistic Regression, Decision Trees, Random Forests, GBDT Regressor, OneVsRestClassifier multi-class solver, and MLPs.
SwiftCluster →
In-memory VectorStore cosine search, SVD-based PCA dimensionality reduction, DBSCAN density clustering, IsolationForest, LocalOutlierFactor, and KMeans++.
SwiftOptimize →
Forecast & Regression Errors Suite (RMSE, MAE, MAPE, R²), ROC-AUC & PR-AUC, AutoML model selection, K-Fold cross-validation, GridSearchCV, and RandomizedSearchCV.
SwiftForecast →
AutoARIMA with zero-variance protection, ARIMA, SARIMA, Holt-Winters exponential smoothing, GARCH volatility, Kalman Filter 1D, STL decomposition, and Koopman operator EDMD.
SwiftNLP →
ChatMessage & ChatTemplate (Llama 3, ChatML, Mistral), VADER Sentiment, NaiveBayes, TextNormalizer, multilingual stopwords, BPETokenizer, NGramTokenizer, and TFIDFVectorizer.
SwiftExplain →
Parallel KernelSHAP, fast TreeSHAP feature attributions, PartialDependencePlot (PDP), and PermutationImportance metrics.
SwiftLLM →
Verified local LLM runtime, SamplingConfiguration logits pipeline, RoPE incremental decoding, two-stage prefill/decode KVCache, QuantizedTensor zero-copy GGUF, Metal MSL GEMV, and MLX.compile.
SwiftVisualization →
Standalone interactive HTML chart exporters: correlation heatmaps, ROC curves, feature importances, and confusion matrix displays.
SwiftVision →
Real YOLOv8n object detection, YOLOv8-Seg instance segmentation with proto masks, CLIP vision-language multimodal projector, and U-Net architecture.
SwiftDatabase →
Direct relational database connectors (SQLite, PostgreSQL, MySQL) with hardened C-pointer lifecycles, explicit connection closing, and zero-copy DataFrame ingestion.
SwiftAgent →
Autonomous ReAct reasoning loop with resilient self-correction, AgentParameterSchema validation, SwiftSciToolbox, SemanticVectorMemory (HNSW), loop sentry, and SwiftUI AgentDialogueController.
SwiftWorkflow →
Production end-to-end ML workflows: training-only imputation, standardization, PCA projection, logistic classification, versioned JSON artifact persistence, and zero test-leakage evaluation.