Releases

Tools for building server-initiated RDMA side-channels

oxicode AI agent SDK — build isolated, multi-agent AI systems

State machine for the QUIC transport protocol

Specialized messaging domain for the Aves async application runtime

Specialized knowledge domain for the Aves async application runtime

Specialized discussion domain for the Aves async application runtime

Typed async application runtime with domain messaging, RPC, and durable workflow automation

DuckBridge: move databases between DuckDB and SQL Server / PostgreSQL / Parquet, translate SQL between dialects, and ship results back as a restorable SQL Server .bak.

Specialized discussion domain for the Aves async application runtime

Specialized knowledge domain for the Aves async application runtime

Typed async application runtime with domain messaging, RPC, and durable workflow automation

Specialized messaging domain for the Aves async application runtime

Core Rust SDK for NeMo Relay observability, scope management, and runtime instrumentation.

Shared serializable data model types for NeMo Relay.

Rust SDK for programmatic control of the GitHub Copilot CLI via JSON-RPC.

Argonauts Code's app controller and run host: the conversation, routing, choosing Captain or crew, and the run host each run has.

Auto-generated Rust SDK for the Alien Platform API

Safe, typed Rust wrappers for NVIDIA FlashInfer's inference-serving kernels: batched paged-KV attention decode, decode-time KV-cache append, cascade / prefix-cache attention-state merge, and sort-free top-K / top-P / min-P sampling. The canonical vLLM-style serving surface for the baracuda CUDA stack. Apache-2.0 (FlashInfer upstream).

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.

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`.

Alien SDK for Rust — app bindings for storage, KV, queues, and vaults

Alien direct in-process resource bindings

The mediaio family in one dependency: the shared media vocabulary, the read and write faces, the editing timeline, and on request the FFmpeg and WebCodecs adapters.

The FFmpeg write adapter of the mediaio family. In this release: the encoder-by-name catalogue, with GPL encoders reachable only behind the `gpl` feature and refused by name without it.

FFmpeg adapter for the `mediadecode` abstraction layer — implements its `VideoAdapter` / `AudioAdapter` / `SubtitleAdapter` traits and the matching push-style decoder traits, with hardware-acceleration auto-probe across VideoToolbox / VAAPI / NVDEC / D3D11VA and software fallback via ffmpeg-next.

The FFmpeg base of the mediaio family: the link to libavcodec and libavformat, and the crossings between FFmpeg's memory and the shared media vocabulary — carriers, packet and frame conversion, codec tickets, limits, and the FFmpeg resampler.

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).

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 NVIDIA cuVS (GPU vector search / approximate nearest neighbours).

Rust SDK for programmatic control of the GitHub Copilot CLI via JSON-RPC.