# Changelog

All notable changes to this project 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).

## [2.0.0-alpha] - 2026-02-14

### Added — Wave 1: Algebra & Symbolic Features
- `poly` — Sparse multivariate polynomials with Lex, DegLex, DegRevLex orderings
- `pattern` — Advanced pattern matching with multi-wildcards, conditions, type filters
- `precision` — Arbitrary precision BigFloat, error propagation, finite fields GF(p)
- `symbols` — Symbol attributes (real, positive, symmetric, antisymmetric, linear, tags)
- `tensor` — Symbolic tensors with Einstein summation, index contraction, Minkowski metric
- `number_theory` — 30+ number theory functions (primes, factorization, CRT, Fibonacci, Bernoulli)
- `units` — Dimensional type system with SI 7-vector
- `discover` — Symbolic regression via genetic programming
- `proof` — Step-by-step proof traces with LaTeX/Markdown export
- `confidence` — Expression confidence scoring
- `transforms` — Laplace and Fourier transforms
- `limits` — Limit computation with 7 strategies (L'Hôpital, Taylor, growth rate, etc.)
- `codegen` — Code generation for 7 targets (Python, C, Rust, Julia, JS, GLSL, WGSL)
- `ode` — First-order ODE solver (separable, linear, exact, Bernoulli)

### Added — Wave 2: Bindings & Streaming
- `symclaw-python` — PyO3 bindings with operator overloading, full API coverage
- `symclaw-collab` — Multi-user WebSocket collaboration server
- `streaming` — Compact binary encoding, file I/O, streaming expression iterators

### Added — Performance Wave
- `modular_gcd` — Zippel sparse modular GCD algorithm
- `packed_poly` — FORM-inspired flat byte-buffer polynomial representation (4-8× less memory)
- `simd_arith` — Montgomery multiplication, batch modular ops, AVX2 acceleration

### Added — Wiring & Integration
- CLI expanded to 20+ commands (`:diff`, `:integrate`, `:limit`, `:codegen`, `:solve`, `:proof`, etc.)
- WASM expanded to 24 browser-ready exports
- Skill expanded to 22 JSON-RPC actions
- Integration tests across all modules
- End-to-end tests for CLI, WASM, and skill workflows

### Added — GPU Wave (World First)
- `symclaw-gpu` crate with 9 GPU modules via CubeCL
- `gpu::poly_gcd` — Batch polynomial evaluation for Zippel GCD (10-100× speedup)
- `gpu::ntt` — Number Theoretic Transform polynomial multiply (100-1000× speedup)
- `gpu::groebner` — F4 Gröbner basis with GPU row reduction (10-100× speedup)
- `gpu::discover` — GPU symbolic regression fitness evaluation (50-100× speedup)
- `gpu::eval` — Batch expression evaluation (10-50× speedup)
- `gpu::linalg` — GPU matrix operations (10-50× speedup)
- `gpu::ode` — Parallel ODE parameter sweeps (10-100× speedup)
- `gpu::monte_carlo` — GPU Monte Carlo integration (50-500× speedup)
- Multi-backend support: CUDA, ROCm, Metal, WebGPU, Vulkan, CPU fallback

### Added — Coverage Boost
- Test count: 426 → 1,173
- Line count: ~11,000 → 39,689
- Module count: ~12 → 48
- Crate count: 4 → 8

### Changed
- Architecture expanded from 3-layer to 8-crate workspace
- AST now uses `Arc<Expr>` throughout for zero-copy sharing
- Simplifier upgraded with e-graph integration and multi-pass pipeline

## [0.1.0] - 2026-02-13

### Added
- Core symbolic engine: parser, simplifier, evaluator, LaTeX renderer
- E-graph equality saturation via egg crate with 30+ rewrite rules
- Symbolic differentiation (all elementary functions, chain rule, higher-order, partial)
- Symbolic integration (power, trig, exp, substitution, by-parts)
- Equation solver (linear, quadratic, polynomial, trig/exp, 2×2 systems)
- Taylor/Maclaurin series expansion with fast-path known functions
- Interactive CLI REPL with Unicode output, variable assignment
- WASM bindings (12 functions)
- OpenClaw skill bridge (JSON-RPC, 9 actions)
- Canvas Manipulate (interactive plotting with parameter sliders)
- Docker deployment (multi-stage build, compose)
- Nix flake (crane builds, dev shell, NixOS module)
- GitHub Actions CI/CD (test, cross-build, release)
- Pi cluster deployment scripts
- 426 tests (unit + property-based)