/ Docs

Changelog

All notable changes to Vectrix are documented here. This project follows Semantic Versioning and the Keep a Changelog format.

[0.0.8] - 2026-03-03

Built-in Rust engine release — Rust acceleration expanded to all engines (13 → 25 functions) and compiled into every wheel. No [turbo] extra needed — pip install vectrix includes the Rust engine like Polars.

Added

Rust Engine Expansion (13 → 25 functions)

  • GARCH: garch_filter, egarch_filter, gjr_garch_filter
  • TBATS: tbats_filter
  • DTSF: dtsf_distances, dtsf_fit_residuals (O(n²) pattern matching — biggest speedup)
  • MSTL: mstl_extract_seasonal, mstl_moving_average
  • Croston: croston_tsb_filter
  • ESN: esn_reservoir_update
  • 4Theta: four_theta_fitted, four_theta_deseasonalize

CI/CD: macOS x86_64 wheel

  • Added macos-13 build target — now 4 platform builds (Linux, macOS ARM, macOS x86, Windows)

Changed

  • All documentation updated: “optional Rust turbo” → “built-in Rust engine”
  • Removed [turbo] extra from all installation guides
  • Landing page rewritten: Hero, Features, Install, Performance sections

[0.0.7] - 2026-03-02

AI integration release — llms.txt, MCP server (10 tools, 2 resources, 2 prompts), Claude Code skills (3).

Added

llms.txt / llms-full.txt

  • llms.txt: Structured project overview following the llms.txt standard with documentation links, quick start, and API sections
  • llms-full.txt: Complete API reference (every class, method, parameter, return type) for full library understanding
  • Deployed to GitHub Pages root and included in PyPI package

MCP Server (Model Context Protocol)

  • 10 tools: forecast_timeseries, forecast_csv, analyze_timeseries, compare_models, run_regression, detect_anomalies, backtest_model, list_sample_datasets, load_sample_dataset
  • 2 resources: vectrix://models, vectrix://api-reference
  • 2 prompts: forecast_workflow, regression_workflow
  • Compatible with Claude Desktop, Claude Code, and any MCP client
  • Install: pip install "vectrix[mcp]"

Claude Code Skills (3)

  • vectrix-forecast: Time series forecasting workflow with full API reference
  • vectrix-analyze: DNA profiling, anomaly detection, regime analysis
  • vectrix-regress: R-style regression, diagnostics, variable selection

[0.0.6] - 2026-03-02

Documentation and deployment release — tutorials, showcases, EasyForecastResult enhancements, unified SvelteKit + MkDocs GitHub Pages.

Added

EasyForecastResult Enhancements

  • compare(): Side-by-side model comparison table with sMAPE, MAPE, RMSE, MAE
  • all_forecasts(): DataFrame of all valid model forecasts
  • Accuracy attributes: .mape, .rmse, .mae, .smape on result objects
  • Vectrix._refitAllModels(): Refit all valid models for compare/all_forecasts

Tutorials (6 topics x 2 languages = 12 files)

  • Quickstart, analyze, regression, models, adaptive, business

Showcases (marimo interactive notebooks)

  • Korean Economy Forecasting, Korean Regression, Model Comparison, Business Intelligence
  • Companion markdown pages for GitHub Pages

Changed

Unified GitHub Pages Deployment

  • SvelteKit landing page at root (/vectrix/)
  • MkDocs documentation at /vectrix/docs/
  • Single docs.yml workflow builds and merges both

[0.0.5] - 2026-03-02

Rust turbo extended — DOT 24x, AutoCES 12x, 4Theta 11x acceleration. 13 total Rust-accelerated functions.

Changed

Rust Turbo Mode Extended (vectrix-core 0.2.0)

  • DOT (Dynamic Optimized Theta): 68ms to 2.8ms (24x faster)
  • AutoCES (Complex Exponential Smoothing): 118ms to 9.6ms (12x faster)
  • 4Theta (Adaptive Theta Ensemble): 63ms to 5.6ms (11x faster)
  • Total: 13 Rust-accelerated functions (was 9, added 4 new)
  • 3-tier fallback preserved: Rust > Numba JIT > Pure Python
  • Bit-identical results with Python reference implementations

[0.0.4] - 2026-03-02

Quality release — English docstrings across 60+ modules, DOT/CES as default model candidates, 573 tests (+186).

Changed

English Docstring Conversion

  • Complete Korean to English conversion across all 60+ source modules
  • All docstrings, error messages, comments now in English
  • API Reference documentation renders correctly in English

Model Selection Improvement

  • DOT and AutoCES now included as default model candidates
  • M4-validated: DOT OWA 0.905, AutoCES OWA 0.927

Added

Test Coverage Expansion (387 to 573, +48%)

  • test_new_models.py: 45 tests for DTSF, ESN, 4Theta
  • test_business.py: 45 tests for anomaly detection, backtesting, metrics, what-if
  • test_infrastructure.py: 43 tests for flat defense, hierarchy, batch, persistence
  • test_engine_utils.py: 53 tests for ARIMAX, cross-validation, decomposition

[0.0.3] - 2026-02-28

Rust turbo mode — 9 accelerated functions, 5-10x speedup, pre-built wheels for all platforms. Built-in sample datasets. pandas 2.x compatibility fixes.

Added

Rust Turbo Mode (vectrix-core)

  • Native Rust extension via PyO3 + maturin
  • 9 accelerated functions: ets_filter, ets_loglik, css_objective, seasonal_css_objective, ses_sse, ses_filter, theta_decompose, arima_css, batch_ets_filter
  • 3-tier fallback: Rust > Numba JIT > Pure Python
  • Pre-built wheels for Linux, macOS (x86 + ARM), Windows
  • Install: pip install "vectrix[turbo]"

Built-in Sample Datasets

  • 7 datasets: airline, retail, stock, temperature, energy, web, intermittent
  • loadSample(name) and listSamples() API

Changed

Performance Improvements

  • AutoETS: 348ms to 32ms (10.8x faster)
  • AutoARIMA: 195ms to 35ms (5.6x faster)
  • Theta: 1.3ms to 0.16ms (8.1x faster)
  • forecast() end-to-end: 295ms to 52ms (5.6x faster)

Fixed

  • pandas 2.x frequency deprecation: "M" to "ME", "Q" to "QE", "Y" to "YE", "H" to "h"

[0.0.2] - 2026-02-28

Foundation models (Chronos, TimesFM), deep learning (NBEATS, NHITS, TFT), VAR/VECM multivariate, multi-country holidays, pipeline system, probabilistic distributions.

Added

Foundation Model Wrappers (Optional)

  • ChronosForecaster: Amazon Chronos-2 zero-shot forecasting
  • TimesFMForecaster: Google TimesFM 2.5 with covariate support

Deep Learning Model Wrappers (Optional)

  • NeuralForecaster: NeuralForecast wrapper for NBEATS, NHITS, TFT
  • Convenience classes: NBEATSForecaster, NHITSForecaster, TFTForecaster

Probabilistic Forecast Distributions

  • ForecastDistribution: Parametric distribution forecasting (Gaussian, Student-t, Log-Normal)
  • DistributionFitter: Automatic distribution selection via AIC
  • empiricalCRPS: Closed-form Gaussian CRPS + Monte Carlo CRPS

Multivariate Models

  • VARModel: Vector AutoRegression with automatic lag selection
  • VECMModel: Vector Error Correction with cointegration rank estimation

Multi-Country Holiday Support

  • US, Japan, China, Korea holidays
  • getHolidays(year) and adjustForecast() API

Pipeline System

  • ForecastPipeline: sklearn-style sequential chaining
  • 8 transformers: Differencer, LogTransformer, BoxCoxTransformer, Scaler, Deseasonalizer, Detrend, OutlierClipper, MissingValueImputer

Changed

  • Parallelized model evaluation via ThreadPoolExecutor
  • 346 tests (up from 275)

[0.0.1] - 2026-02-27

Initial release — 30+ forecasting models, adaptive intelligence, Easy API, regression suite, business intelligence, 275 tests.

Added

Core Forecasting Engine (30+ Models)

  • AutoETS, AutoARIMA, Theta, DOT, AutoCES, AutoTBATS, GARCH/EGARCH/GJR-GARCH
  • Croston (Classic/SBA/TSB/Auto), Logistic Growth, AutoMSTL
  • Baseline: Naive, Seasonal Naive, Mean, Random Walk with Drift, Window Average

Novel Methods

  • Lotka-Volterra Ensemble, Phase Transition Forecaster, Adversarial Stress Tester
  • Hawkes Intermittent Demand, Entropic Confidence Scorer

Adaptive Intelligence

  • Regime Detection (HMM), Self-Healing Forecast (CUSUM + EWMA)
  • Constraint-Aware Forecasting (8 constraint types)
  • Forecast DNA (65+ features, meta-learning model recommendation)
  • Flat Defense (4-level diagnostic/detection/correction/prevention)

Easy API

  • forecast(): One-call forecasting with automatic model selection
  • analyze(): Time series DNA profiling and anomaly identification
  • regress(): R-style formula regression with full diagnostics
  • quick_report(): Combined analysis + forecast report

Regression and Diagnostics

  • 5 methods: OLS, Ridge, Lasso, Huber, Quantile
  • Full diagnostics: Durbin-Watson, Breusch-Pagan, VIF, Jarque-Bera
  • Variable selection: Stepwise, regularization CV, best subset
  • Time series regression: Newey-West HAC, Cochrane-Orcutt, Prais-Winsten

Business Intelligence

  • Anomaly detection, what-if analysis, backtesting
  • Hierarchy reconciliation (Bottom-up, Top-down, MinTrace)
  • Prediction intervals (Conformal + Bootstrap)

Infrastructure

  • Batch forecasting with parallel execution
  • Model persistence (.fxm format)
  • TSFrame, Global model, Numba JIT acceleration
  • GitHub Actions CI (Python 3.10-3.13, Ubuntu + Windows)
  • PyPI trusted publisher deployment