The inside story of building the largest pure-Rust sovereignty stack
159 posts
OxigenAI is a Pure Rust GovTech reimplementation of the Digital Agency's government legal-AI system (源内 / GenAI). It keeps the familiar law-report API but adds OxiZ SMT contradiction detection and Legalis-RS LegalResult<T> outcome classification — turning a probabilistic LLM+RAG service into a verifiable computational-law platform.
A maintenance release for OxiHuman, the privacy-first pure Rust parametric human body generator that runs entirely client-side via WASM/WebGPU. 0.1.2 makes the MakeHuman-backed test suite fully portable — dataset and asset roots now resolve through MAKEHUMAN_DATA_DIR and OXIHUMAN_ASSETS_DIR, and fixtures skip gracefully when unset. ~943,000 lines of Rust, 32,791 passing tests, zero hard-coded paths.
OxiLLaMa 0.1.3 is the Pure Rust LLM inference engine and sovereign alternative to llama.cpp. This release adds BLOOM + Phi-3.5-MoE architectures (now 27 total), a 5-stage advanced sampler suite (DRY/XTC/TypicalP/TopA/Eta) that is byte-identical at defaults, embedding pooling modes, a drop-in /v1/responses API with per-API-key rate limiting, AVX-512 IQ kernels at ~2x per-iteration throughput, GPU-resident sampling kernels, and zero-copy DLPack PyTorch interop — 2,461 tests passing.
OxiRS 0.3.0 turns the Rust-native Semantic Web stack enterprise-ready: SOC2/GDPR audit logging, OIDC + SAML 2.0 single sign-on, a HuggingFace/Ollama/GGUF model marketplace, LoRA fine-tuning for GraphRAG, and published LTS + enterprise support policies.
OxiBonsai 0.1.3 makes sub-2-bit serving smarter: a prefix-cache-aware engine that reuses KV-cache across requests with byte-identical cold/warm parity, runtime tokenizer auto-detection, and a GPU weight cache that uploads once. Sub-2-bit Pure Rust sovereign AI inference for the COOLJAPAN ecosystem.
A focused OxiCode patch: a deny.toml policy gate to ban unsafe or incompatible crates, the bincode dev-dependency pinned to =2.0.1, OxiARC pure Rust compression bumped to 0.2.7, and more robust temp-file handling in the file-I/O test suite. Same 100% bincode-2.0 binary compatibility, same Pure Rust guarantee.
The pure-Rust NVIDIA CUDA Toolkit replacement adds nine new GPU deep-learning crates — generative diffusion, graph neural nets, Mamba SSMs, vision transformers, audio/speech, time-series, Bayesian DL, federated learning, and NAS — growing to ~320K lines across 37 crates with 9,568 passing tests. No CUDA SDK, no nvcc.
OxiLean 0.1.2 lands real SMT solving via OxiZ, a complete 157-instruction WebAssembly bytecode interpreter wired to the real kernel/parse/elab pipeline, keccak256-correct EVM/Solidity ABI selectors, real Gröbner-basis reduction for polyrith, and 33,091 passing tests — all in a Pure Rust theorem prover.
SciRS2 is the pure-Rust SciPy/scikit-learn replacement. 0.4.3 introduces the brand-new scirs2-symbolic crate (symbolic differentiation + algebraic simplification), full WASM TypeScript declarations & React hooks, an inverse wavelet packet transform, and modernized dependencies — 34,883 tests, 2.94M lines, 32 crates, zero warnings.
Kizzasi 0.2.1 turns the Pure-Rust AGSP into a full training + deployment stack: full backprop through SSM recurrence, RWKV v5/v7, Neural ODE and Spiking models, LoRA fine-tuning, GGUF loading, gRPC/REST servers, Python bindings, and no_std embedded inference.
ToRSh is a pure-Rust, PyTorch-compatible deep-learning framework with native tensor sharding. 0.1.2 lands real AVX2/NEON SIMD for f32 ops and activations, a true zero-copy buffer pool (100% heap-block reduction on hot loops), and SIMD + parallel enabled by default.
OxiMedia 0.1.6 — patent-free, memory-safe FFmpeg + OpenCV replacement in pure Rust. This release resolves 13 codec stubs (accel color conversion, Vorbis VQ decode, ACES ODT variants), upgrades to OxiFFT 0.3.0 for a ~4x faster DCT-II, and splits the EXR reader into 9 modules. 108 crates, 81,582 tests, zero C/Fortran in default builds.
OxiZ is a pure-Rust SMT solver and Z3 replacement. 0.2.1 sharpens the EUF (congruence-closure) hot path: reusable allocation buffers, O(k) incremental pop() via a sig_table/fingerprint_table trail, cache-friendly ENode layout, and new production EUF criterion benchmarks. 100% Z3 parity, still pure Rust.
OxiLLaMa 0.1.2 is the Pure Rust LLM inference engine and sovereign alternative to llama.cpp. This release adds `oxillama hub pull/list/rm` (hf-hub, no Python), a full-screen TUI chat with live streaming (ratatui), conversation save/resume serialized via oxicode with SHA-256 integrity, and real weight loading for DBRX, Grok-1, and Mamba-2.
OxiLLaMa is a Pure Rust LLM inference engine — the sovereign alternative to llama.cpp. Version 0.1.1 ships a tiled FlashAttention CPU kernel, true continuous batching with zero padding waste, fused dequant+GEMM (~12% Q4_K_M decode gain), 5 new architectures (DBRX, Grok-1, Mamba-2, DeepSeek-V3, and more), and GPU coverage extended to 10 quantization types.
Pure-Rust legal statute engine: 0.1.5 is a sovereignty and hardening release. The last non-pure-Rust dependency (printpdf) is replaced by the COOLJAPAN fop-render PDF backend, an RSA CVE is eliminated, every unwrap() is removed from production code, and the default build is now 100% Pure Rust across 23 jurisdictions. 14,763 tests passing.
Pure Rust archive and compression with a new in-place editing workflow. OxiArc 0.2.7 adds the `oxiarc add` command for appending files to existing ZIP/TAR/LZH archives, lenient recovery of malformed archives, async LZH/TAR streaming, cooperative cancellation, colored output, and man pages. 12 formats, 10 codecs, 12 crates, ~47k SLoC, 1,041 tests.
Pure-Rust FFmpeg+OpenCV replacement: OxiMedia 0.1.5 adds the oximedia-ml crate — typed ML pipelines (SceneClassifier, ShotBoundaryDetector, and more) on the Pure-Rust OxiONNX runtime, with a Python oximedia.ml submodule and an opt-in, symbol-free-by-default design. Plus a codec-decoder honesty pass. 108 crates, ~2.68M SLoC, 81,383 tests.
OxiMedia is the pure-Rust FFmpeg + OpenCV replacement. Release 0.1.4 wires MJPEG and APV codecs end-to-end, adds OpenDML AVI and animated JPEG-XL containers, low-latency DASH/CMAF streaming, deeper FFmpeg CLI compat (-crf/-vf/two-pass), and fixes the JPEG encoder (round-trip PSNR 6.16 -> 32.53 dB). 108 crates, ~2.67M SLoC, 81,150+ tests passing.
OxiBonsai 0.1.2 adds ONNX ingestion: pull an onnx-community Ternary ONNX release (MatMulNBits, bits=2) and repack it straight to OxiBonsai's GGUF TQ2_0_g128 with a single command — driven by the pure-Rust oxionnx-proto reader, no Python and no onnxruntime. Sub-2-bit sovereign AI inference for the COOLJAPAN ecosystem.
OxiGDAL 0.1.4 lands Weiler-Atherton polygon clipping, Karney geodesic area, DE-9IM topology predicates, and marching-squares contours, plus AVX2/NEON SIMD resampling, a completed COPC point-cloud reader, R-tree STR bulk loading + kNN, and a migration from ort to oxionnx for fully Pure Rust ONNX inference. 76 crates, 12,064 passing tests.
Five days after its 1-bit debut, OxiBonsai grows GPUs: a native CUDA NVRTC backend (~21.9 tok/s on Ternary-Bonsai-1.7B, RTX 3060) and a fused Metal full-forward path (~50 tok/s, ~13x speedup) — plus the new ternary TQ2_0_g128 quant family, with NEON/AVX2/AVX-512 GEMV so it flies on CPU too. Sub-2-bit Pure Rust sovereign AI inference for the COOLJAPAN ecosystem, still with no llama.cpp, no BLAS, no C/Fortran.
A small maintenance release for OxiCUDA, the pure-Rust replacement for the NVIDIA CUDA Toolkit. Workspace-wide documentation and quality improvements, with all 28 crates aligned to 0.1.4 so the stack ships in lockstep. The only runtime dependency is the NVIDIA driver.
A quality-and-docs maintenance release for the pure-Rust NVIDIA CUDA Toolkit replacement — workspace-wide polish, internal version alignment to 0.1.3, and continued growth to ~260K lines of safe Rust across 28 crates. The only runtime dependency is still the NVIDIA driver.