Releases

RLX multi-node pipeline-parallel execution: graph partition into weight-sharded stages + tensor transport, to pool RAM across machines for models larger than any single node.

C ABI for the RLX distributed node — link a mesh worker into iOS, an embedded host, or any non-Rust shell

Hidden-vetter admission (ZKP) on the Predicate Credential System: the community's issuer and verifier, and the member-side vetter and applicant engines a VTA runs

Device-free static graph checker for RLX — backend dispatch, fusion, shape/dtype and numeric diagnostics (`cargo rlx check`)

VTA TEE bootstrap — Nitro/SEV-SNP attestation providers, KMS attest/decrypt + storage-key derivation, the anchor MAC, and first-boot DID autogen

WebAssembly entry point for RLX — run models in the browser (CPU backend; WebGPU device bring-up)

Background TTL sweepers for the VTA's core keyspaces — ACL grant expiry, pending-consent expiry, and soft-deleted-vault purge

Black-box optimization and FMQ/QGBS search (trust-region Q-guidance, arXiv:2605.12416).

PyTorch (torch.export) → RLX HIR import: aten→rlx op registry, runnable bundle + generated crate

ONNX → RLX HIR import (bundle loader, ONNX protobuf, op lowering)

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.

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).

WebVH hosting infrastructure for the VTA — the local did:webvh DID-record and hosting-server store

Native ndarray alternative for RLX — NumPy-style tensors that trace, fuse, and run on any backend via the rlx compiler

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

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`).

VTA holder credential vault — storage, query, receive/verify, present, and status refresh for credentials a holder stores on a VTA

RLX runtime — feature-gated backends, session API, compile+execute pipeline

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

VTA policy subsystem — the regorus (Rego) engine, the default policy bundle, consent model, and decision evaluators

VTA persona store — the holder's own identity attributes, the profiles that project over them, the bindings that assign a profile to a persona DID, and the contacts peers disclose

VTA external accounts — the VTA as the key authority for cloud and third-party accounts: account store, auth-model drivers, downscoped credential issuance

Shared mid-layer VTA services — trust-context storage, the sealed-transfer seal helper, and the sealed-bootstrap nonce store

VTA backup/restore subsystem — encrypted full-state export/import, the sealed backup-bundle store, and its TTL sweeper

VTA key management — master-seed storage, BIP-32 key derivation, key wrapping, and the seed-store backend selection

HyperChad FLTK renderer package

HyperChad egui renderer package