Releases

Unified ML op facade for the baracuda CUDA ecosystem. Exposes every primitive an ML framework would expect (union of PyTorch torch.* + nn.functional and JAX lax.* / numpy ops) through a single Plan-based Rust surface, internally dispatching to baracuda-cutlass, the baracuda-* NVIDIA-library wrappers, or bespoke baracuda-kernels-sys kernels.

Safe Rust wrapper for compiled CUTLASS kernels: plan-based GEMM and grouped GEMM with caller-supplied workspace, typed device-buffer arguments, and capture-safe launch.

Safe Rust wrapper for baracuda's clean-fork of Hiroyuki Ootomo's ozIMMU — Ozaki-scheme FP64 GEMM that synthesizes a DGEMM from S^2 int8 tensor-core matmuls. Provides an RAII handle + drop-in `dgemm` shim suitable for the `BackendKind::Ozaki` path of `baracuda-kernels`'s `GemmPlan`. Opt-in (NOT bit-equivalent to native DGEMM); the default FP64 path stays on CUTLASS / cuBLAS.

Pure Rust Pinget core library that works directly with source caches, REST endpoints, and installed package state without COM.

Idiomatic Rust wrappers for the NVIDIA CUDA stack (Driver API, Runtime API, NVRTC, cuBLAS, cuDNN, NCCL, NVML, ...). Umbrella crate.

Megatron-LM-style tensor-parallel primitives (Column / Row Parallel Linear) for the baracuda CUDA stack. Pure-composition crate — local GEMM via baracuda-cublas + cross-rank collectives via baracuda-nccl. No new CUDA kernels. NEW in Phase 57; deliberate scope expansion (distributed-training-framework-adjacent). Off-by-default in baracuda-kernels via the `megatron_tp` cargo feature so non-distributed consumers don't pay the dep surface cost. Algorithmic reference: Shoeybi et al. arXiv:1909.08053 (NVIDIA Megatron-LM, Apache-2.0).

Safe Rust wrappers for the CUDA Runtime API (devices, streams, events, managed memory, kernel launch via the library API).

Build + raw FFI bindings to baracuda's clean-fork of Hiroyuki Ootomo's ozIMMU — the Ozaki-scheme FP64 GEMM library that synthesizes a DGEMM from S² int8 tensor-core matmuls. Phase 44b internalized the upstream sources under `cuda/` (no more `vendor/` subdir; cutf submodule eliminated). Linked statically into the baracuda CUDA stack; consumed by the safe wrapper crate `baracuda-ozimmu`. MIT-licensed (original ozIMMU MIT — see `ATTRIBUTION.md`).

Safe Rust wrappers for NVIDIA NVRTC (compile CUDA C++ to PTX at runtime).

Safe Rust wrappers for NVIDIA NPP (Performance Primitives). Core + signal subset at v0.1.

Embedded JavaScript engine: Deno runtime with Node compatibility over the app's node_modules

An actor-first distributed database

DuckDB adapter for the Quarb query engine

Dev server: on-demand compile over HTTP, WebSocket HMR channel, file watcher

Fork of deno_napi 0.190.0 maintained for oj, carrying oj's finalizer-registry performance patch (Vec -> BTreeMap). Use upstream deno_napi unless you need this patch.

Fork of deno_runtime 0.267.0 maintained for oj, swapping deno_napi for oj_deno_napi (finalizer-registry performance patch). Use upstream deno_runtime unless you need this patch.

Fork of deno_snapshots 0.74.0 maintained for oj: the v8 snapshot can be consumed from a prebuilt per-target bundle (OJ_SNAPSHOT_ARCHIVE) for reproducible builds. Use upstream deno_snapshots unless you need this.

oj's fork of deno_process: child process groups + spawn hook

In-process transport: composes a kernel directly in the host process.

Safe Rust wrappers for NVIDIA cuVS (GPU vector search / approximate nearest neighbours).

Safe Rust wrappers for NVIDIA cuSOLVER (dense LU factorization at v0.1).

Safe Rust wrappers for NVIDIA cuFFT.

SurrealDB vector store implementation for the rig framework

Build + raw FFI bindings to baracuda's port of NVIDIA TransformerEngine's FP8 cast/transpose + delayed-scaling recipe primitives. Cast/recipe subset only — `normalization` / `fused_rope` / `fused_attn` / `fused_softmax` / `activation` / `gemm` deliberately skipped (overlap existing baracuda Phase 3/5/14/17/30/31/36/41/42). NO cuDNN dep (recipe + cast paths don't need it; `fused_attn` would, and we skip it); NO pybind11 (the safe wrapper lives in `baracuda-transformer-engine` and exposes a raw C ABI defined in `csrc/baracuda_te_shim.cu`). Apache-2.0 per upstream — see `ATTRIBUTION.md`.

The intray as a tuplespace: urn:space:{name} — Sink drops a tuple (out), Source lists/reads/filters (rd), Delete atomically takes one (in), associative match via SPARQL ASK. A physical, capability-gated space; later slices add the reactive engine and encrypt-on-drop.

Raw FFI bindings and dynamic loader for NVIDIA nvJPEG.

Raw C-ABI FFI surface for the vendored FlashInfer inference kernels (paged-KV decode, paged-KV append, cascade state merge, sort-free sampling). The launcher shims + vendored FlashInfer sources are compiled by `baracuda-kernels-sys`; this crate is a thin re-export facade so the FlashInfer C-ABI is reachable under its own crate name. Apache-2.0 (FlashInfer upstream) — see `crates/baracuda-kernels-sys/vendor/flashinfer/`.

Safe Rust wrappers for the CUDA Driver API (devices, contexts, streams, events, memory, kernels, graphs).

Raw FFI bindings and dynamic loader for NVIDIA cuVS (GPU vector search / ANN).

Safe Rust wrappers for CUPTI. Scaffolding at v0.1.