2026-03-03 Built-in Rust engine: 25 accelerated functions, no [turbo] extra needed

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)
- MSTL: mstl_extract_seasonal, mstl_moving_average
- Croston: croston_tsb_filter
- ESN: esn_reservoir_update
- 4Theta: four_theta_fitted, four_theta_deseasonalize

Messaging overhaul: "optional Rust turbo" → "built-in Rust engine"
- README, landing page, docs, SEO metadata, llms.txt all updated
- CI: added macOS x86_64 (macos-13) wheel build

Author: eddmpython

.github/workflows/publish.yml

@@ -38,13 +38,21 @@ jobs:
         run: pytest tests/ -v --tb=short
 
   build-wheels:
-    name: Build wheels (${{ matrix.os }})
+    name: Build wheels (${{ matrix.os }} ${{ matrix.target }})
     needs: test
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-14, windows-latest]
+        include:
+          - os: ubuntu-latest
+            target: ""
+          - os: macos-14
+            target: aarch64-apple-darwin
+          - os: macos-13
+            target: x86_64-apple-darwin
+          - os: windows-latest
+            target: ""
 
     steps:
       - uses: actions/checkout@v4
@@ -61,7 +69,7 @@ jobs:
       - name: Build wheels
         uses: PyO3/maturin-action@v1
         with:
-          target: ${{ matrix.os == 'macos-14' && 'aarch64-apple-darwin' || '' }}
+          target: ${{ matrix.target }}
           args: --release --out dist -i python3.10 -i python3.11 -i python3.12 -i python3.13
           sccache: true
           manylinux: auto
@@ -71,7 +79,7 @@ jobs:
       - name: Upload wheels
         uses: actions/upload-artifact@v4
         with:
-          name: wheels-${{ matrix.os }}
+          name: wheels-${{ matrix.os }}-${{ matrix.target || 'default' }}
           path: dist
 
   build-sdist:

CHANGELOG.md

@@ -5,6 +5,39 @@ All notable changes to Vectrix will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.0.8] - 2026-03-03
+
+Built-in Rust engine release — Rust acceleration expanded to all engines and compiled into every wheel. No `[turbo]` extra, no flags — `pip install vectrix` includes the Rust engine like Polars.
+
+### Added
+
+**Rust Engine Expansion (13 → 25 functions)**
+- `garch_filter`, `egarch_filter`, `gjr_garch_filter`: GARCH family negative log-likelihood hot loops
+- `tbats_filter`: Fourier harmonic state update loop
+- `dtsf_distances`, `dtsf_fit_residuals`: O(n²) sliding window pattern matching (biggest speedup)
+- `mstl_extract_seasonal`, `mstl_moving_average`: MSTL seasonal decomposition
+- `croston_tsb_filter`: Croston TSB SES update loop
+- `esn_reservoir_update`: Echo State Network reservoir state computation O(n×N²)
+- `four_theta_fitted`, `four_theta_deseasonalize`: 4Theta fitted values + seasonal decomposition
+- All 25 functions compiled into the default wheel — no separate `[turbo]` install needed
+
+**CI/CD: macOS x86_64 wheel**
+- Added `macos-13` build target for Intel Mac users
+- Now 4 platform builds: Linux x86_64, macOS ARM, macOS x86_64, Windows x86_64
+
+### Changed
+
+**Messaging: "optional Rust turbo" → "built-in Rust engine"**
+- `pyproject.toml`: Description updated to reflect built-in Rust engine
+- `README.md`: Removed all "optional" Rust language, comparison table updated
+- Landing page: Hero, Features, Install, Performance, Numbers sections rewritten
+- Docs (EN/KO): Installation guides rewritten — no `[turbo]` extra mentioned
+- SEO metadata: All "Rust turbo" → "built-in Rust engine" across meta tags, schema, OG tags
+
+**Version sync**: pyproject.toml, Cargo.toml, __init__.py all at 0.0.8
+
+[0.0.8]: https://github.com/eddmpython/vectrix/compare/v0.0.7...v0.0.8
+
 ## [0.0.7] - 2026-03-02
 
 AI integration release — llms.txt for instant AI understanding, MCP server for tool use, Claude Code skills for workflow automation.

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "vectrix"
-version = "0.0.7"
+version = "0.0.8"
 edition = "2021"
 license-file = "LICENSE"
 

README.md

@@ -10,12 +10,12 @@
 
 <br>
 
-<h3>Pure Python Time Series Forecasting Engine</h3>
+<h3>Time Series Forecasting Engine — Built-in Rust Acceleration</h3>
 
 <p>
 <img src="https://img.shields.io/badge/3-Dependencies-818cf8?style=for-the-badge&labelColor=0f172a" alt="Dependencies">
-<img src="https://img.shields.io/badge/Pure-Python-6366f1?style=for-the-badge&labelColor=0f172a" alt="Pure Python">
-<img src="https://img.shields.io/badge/Rust-Turbo%20Mode-e45a33?style=for-the-badge&labelColor=0f172a&logo=rust&logoColor=white" alt="Rust Turbo">
+<img src="https://img.shields.io/badge/Built--in-Rust%20Engine-e45a33?style=for-the-badge&labelColor=0f172a&logo=rust&logoColor=white" alt="Built-in Rust Engine">
+<img src="https://img.shields.io/badge/Python-3.10+-6366f1?style=for-the-badge&labelColor=0f172a&logo=python&logoColor=white" alt="Python 3.10+">
 </p>
 
 <p>
@@ -50,7 +50,7 @@
 
 ## ◈ What is Vectrix?
 
-Vectrix is a time series forecasting library that runs on **3 dependencies** (NumPy, SciPy, Pandas) with **no compiled extensions**. No C compiler, no cmdstan, no system packages — `pip install` and it works.
+Vectrix is a time series forecasting library with a **built-in Rust engine** for blazing-fast performance. 3 dependencies (NumPy, SciPy, Pandas), no compiler needed — `pip install vectrix` and the Rust-accelerated engine is included in the wheel.
 
 ### Forecasting
 
@@ -103,18 +103,18 @@ A pure-numpy HMM (Baum-Welch + Viterbi) detects regime shifts. When a regime cha
 
 Bottom-up, top-down, and MinTrace reconciliation for hierarchical time series.
 
-### Rust Turbo Mode
+### Built-in Rust Engine
 
-Install `vectrix[turbo]` to unlock Rust-accelerated core loops. No Rust compiler needed — pre-built wheels for Linux, macOS (x86 + ARM), and Windows.
+Every `pip install vectrix` includes a pre-built Rust extension — like Polars, no compiler needed. 25 core hot loops are Rust-accelerated across all forecasting engines.
 
-| Component | Without Turbo | With Turbo | Speedup |
-|:----------|:-------------|:-----------|:--------|
+| Component | Python Only | With Rust | Speedup |
+|:----------|:-----------|:----------|:--------|
 | `forecast()` 200pts | 295ms | **52ms** | **5.6x** |
 | AutoETS fit | 348ms | **32ms** | **10.8x** |
-| AutoARIMA fit | 195ms | **35ms** | **5.6x** |
+| DOT fit | 240ms | **10ms** | **24x** |
 | ETS filter (hot loop) | 0.17ms | **0.003ms** | **67x** |
 
-Turbo is fully optional. Without it, Vectrix falls back to Numba JIT (if available) or pure Python. Your code doesn't change — just install and it's faster.
+Pre-built wheels for Linux (x86_64), macOS (ARM + x86), and Windows. The Rust engine is included in the default installation — no extras, no flags, no `[turbo]`.
 
 ### Built-in Sample Datasets
 
@@ -129,9 +129,9 @@ result = forecast(df, date="date", value="passengers", steps=12)
 
 Available: `airline`, `retail`, `stock`, `temperature`, `energy`, `web`, `intermittent`
 
-### Everything is Pure Python
+### Minimal Dependencies, Maximum Performance
 
-All of the above — forecasting models, regime detection, regression diagnostics, constraint enforcement, hierarchical reconciliation — is implemented in pure Python with only NumPy, SciPy, and Pandas. No compiled extensions, no system dependencies. Rust turbo is optional and never required.
+All of the above — forecasting models, regime detection, regression diagnostics, constraint enforcement, hierarchical reconciliation — runs on just NumPy, SciPy, and Pandas. The Rust engine is compiled into the wheel and loaded automatically. No system dependencies, no compiler, no extra install steps.
 
 <br>
 
@@ -156,16 +156,16 @@ result.plot()
 
 | | Vectrix | statsforecast | Prophet | Darts |
 |:--|:--:|:--:|:--:|:--:|
-| **Pure Python (no C/Fortran)** | ✅ | ❌ (numba) | ❌ (cmdstan) | ❌ (torch) |
-| **Optional Rust acceleration** | ✅ (5-10x) | ❌ | ❌ | ❌ |
+| **Built-in Rust engine** | ✅ (5-67x) | ❌ | ❌ | ❌ |
+| **No compiler needed** | ✅ | ❌ (numba) | ❌ (cmdstan) | ❌ (torch) |
 | **Dependencies** | 3 | 5+ | 10+ | 20+ |
 | **Auto model selection** | ✅ | ✅ | ❌ | ❌ |
 | **Flat-line defense** | ✅ | ❌ | ❌ | ❌ |
 | **Business constraints** | 8 types | ❌ | ❌ | ❌ |
 | **Built-in regression** | R-style | ❌ | ❌ | ❌ |
 | **Sample datasets** | 7 built-in | ❌ | ❌ | ✅ |
 
-> **Comparison notes**: Dependencies counted as direct `pip install` requirements (not transitive). "Pure Python" means the core library runs without compiled extensions — Vectrix's optional Rust turbo (`vectrix[turbo]`) is a separate, opt-in wheel. statsforecast requires Numba JIT compilation; Prophet requires CmdStan (C++ compiler); Darts requires PyTorch. Feature comparison based on statsforecast 2.0+, Prophet 1.1+, Darts 0.31+.
+> **Comparison notes**: Dependencies counted as direct `pip install` requirements (not transitive). Vectrix's Rust engine is compiled into the wheel (like Polars) — no separate install needed. statsforecast requires Numba JIT compilation; Prophet requires CmdStan (C++ compiler); Darts requires PyTorch. Feature comparison based on statsforecast 2.0+, Prophet 1.1+, Darts 0.31+.
 
 <br>
 
@@ -259,9 +259,7 @@ result.plot()
 ## ◈ Installation
 
 ```bash
-pip install vectrix                # Core (numpy + scipy + pandas)
-pip install "vectrix[turbo]"       # + Rust acceleration (5-10x speedup)
-pip install "vectrix[numba]"       # + Numba JIT (2-5x speedup)
+pip install vectrix                # Rust engine included — no extras needed
 pip install "vectrix[ml]"          # + LightGBM, XGBoost, scikit-learn
 pip install "vectrix[all]"         # Everything
 ```
@@ -430,8 +428,8 @@ vectrix/
 ├── global_model/          Cross-series forecasting
 └── datasets.py            7 built-in sample datasets
 
-rust/                         Optional Rust acceleration (vectrix-core)
-└── src/lib.rs             ETS, ARIMA, Theta, SES hot loops (PyO3)
+rust/                         Built-in Rust engine (25 accelerated functions)
+└── src/lib.rs             ETS, ARIMA, DOT, CES, GARCH, DTSF, ESN, 4Theta (PyO3)
 ```
 
 <br>

docs/getting-started/installation.ko.md

@@ -19,33 +19,34 @@
     uv add vectrix
     ```
 
+Rust 엔진이 **wheel에 내장**되어 있습니다 — 별도 설치, 컴파일러 불필요. Polars처럼 설치만 하면 바로 빠릅니다.
+
 ## 선택적 추가 기능
 
 ```bash
-pip install "vectrix[turbo]"       # Rust 가속 (5-10배 속도 향상, Rust 컴파일러 불필요)
-pip install "vectrix[numba]"       # Numba JIT 가속 (2-5배 속도 향상)
-pip install "vectrix[ml]"          # LightGBM, XGBoost, scikit-learn
-pip install "vectrix[foundation]"  # Amazon Chronos-2, Google TimesFM 2.5
-pip install "vectrix[tutorials]"   # 인터랙티브 marimo 튜토리얼
-pip install "vectrix[all]"         # 전체
+pip install vectrix                  # 30+ 모델 + Rust 엔진 내장
+pip install "vectrix[ml]"            # + LightGBM, XGBoost, scikit-learn
+pip install "vectrix[foundation]"    # + Amazon Chronos-2, Google TimesFM 2.5
+pip install "vectrix[all]"           # 전체
 ```
 
-## Rust Turbo Mode
+## 내장 Rust 엔진
 
-`turbo` 옵션은 Rust로 컴파일된 네이티브 확장 `vectrix-core`를 설치합니다. 핵심 예측 루프가 5-10배 빨라집니다. 사전 빌드 wheel 제공:
+25개 핵심 예측 핫 루프가 Rust로 가속되어 모든 wheel에 컴파일됩니다:
 
 - Linux (x86_64, manylinux)
-- macOS (x86_64 + Apple Silicon ARM)
+- macOS (Apple Silicon ARM + x86_64)
 - Windows (x86_64)
 - Python 3.10, 3.11, 3.12, 3.13
 
 Rust 컴파일러 불필요. 코드 변경 없이 설치만 하면 자동으로 빨라집니다.
 
-| 구성요소 | Turbo 없음 | Turbo 적용 | 속도 향상 |
-|:---------|:----------|:----------|:---------|
+| 구성요소 | Python Only | Rust 적용 | 속도 향상 |
+|:---------|:-----------|:----------|:---------|
 | `forecast()` 200pts | 295ms | **52ms** | **5.6x** |
 | AutoETS fit | 348ms | **32ms** | **10.8x** |
-| AutoARIMA fit | 195ms | **35ms** | **5.6x** |
+| DOT fit | 240ms | **10ms** | **24x** |
+| ETS filter (핫 루프) | 0.17ms | **0.003ms** | **67x** |
 
 ## 설치 확인
 
@@ -57,6 +58,12 @@ result = forecast([100, 120, 130, 115, 140], steps=3)
 print(result.predictions)
 ```
 
+Rust 엔진 활성화 확인:
+
+```python
+print(vectrix.TURBO_AVAILABLE)  # True
+```
+
 ## 핵심 의존성
 
 Vectrix는 3개의 필수 의존성만 있습니다:

docs/getting-started/installation.md

@@ -19,33 +19,34 @@
     uv add vectrix
     ```
 
+The Rust engine is **built into the wheel** — no extras, no compiler needed. Like Polars, just install and it's fast.
+
 ## Optional Extras
 
 ```bash
-pip install "vectrix[turbo]"       # Rust acceleration (5-10x speedup, no Rust compiler needed)
-pip install "vectrix[numba]"       # Numba JIT acceleration (2-5x speedup)
-pip install "vectrix[ml]"          # LightGBM, XGBoost, scikit-learn
-pip install "vectrix[foundation]"  # Amazon Chronos-2, Google TimesFM 2.5
-pip install "vectrix[tutorials]"   # Interactive marimo tutorials
-pip install "vectrix[all]"         # Everything
+pip install vectrix                  # 30+ models + built-in Rust engine
+pip install "vectrix[ml]"            # + LightGBM, XGBoost, scikit-learn
+pip install "vectrix[foundation]"    # + Amazon Chronos-2, Google TimesFM 2.5
+pip install "vectrix[all]"           # Everything
 ```
 
-## Rust Turbo Mode
+## Built-in Rust Engine
 
-The `turbo` extra installs `vectrix-core`, a Rust-compiled native extension that accelerates core forecasting loops by 5-10x. Pre-built wheels are available for:
+25 core forecasting hot loops are Rust-accelerated and compiled into every wheel:
 
 - Linux (x86_64, manylinux)
-- macOS (x86_64 + Apple Silicon ARM)
+- macOS (Apple Silicon ARM + x86_64)
 - Windows (x86_64)
 - Python 3.10, 3.11, 3.12, 3.13
 
 No Rust compiler is needed. The acceleration is transparent — your code doesn't change, it just runs faster.
 
-| Component | Without Turbo | With Turbo | Speedup |
-|:----------|:-------------|:-----------|:--------|
+| Component | Python Only | With Rust | Speedup |
+|:----------|:-----------|:----------|:--------|
 | `forecast()` 200pts | 295ms | **52ms** | **5.6x** |
 | AutoETS fit | 348ms | **32ms** | **10.8x** |
-| AutoARIMA fit | 195ms | **35ms** | **5.6x** |
+| DOT fit | 240ms | **10ms** | **24x** |
+| ETS filter (hot loop) | 0.17ms | **0.003ms** | **67x** |
 
 ## Verify
 
@@ -57,6 +58,12 @@ result = forecast([100, 120, 130, 115, 140], steps=3)
 print(result.predictions)
 ```
 
+Check that Rust engine is active:
+
+```python
+print(vectrix.TURBO_AVAILABLE)  # True
+```
+
 ## Core Dependencies
 
 Vectrix has only 3 required dependencies:

landing/src/content/en/changelog.md

@@ -6,6 +6,32 @@ title: Changelog
 
 All notable changes to Vectrix are documented here. This project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html) and the [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) 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).