Packages
Browse packages indexed by the registry.
Rust bindings for atomicparsley
AtomicRingBuffer is a constant-size almost lock-free concurrent ring buffer
Basic utils for concurrent programming. Backoff, spinlocks, seqlock, atomic type wrappers.
A Rust library for thread-safe shared slices that are just about as fast as possible to read while also being writable.
A place to play with monotonic atomic numbers
Atomic file-writes.
Generic, convenient and lock-free `std` atomics via `Atomic<T>`. Can be used with many primitive types (including floats) and with custom types.
Helper crate for `atomig`. Do not use directly, but only through main crate. This helper does not follow semantic versioning!
Derive macros for atomiq.
Macros for the atomize crate
Native Rust runtime for actor-based concurrent and distributed systems. Umbrella crate — pull in subsystems via feature flags.
Backend-agnostic compute-acceleration core. Defines the AccelBackend trait, AccelRef<T> typed pointers, AccelError enum, and CompletionStrategy — the abstraction layer that lets atomr-accel-cuda (NVIDIA), and future ROCm / Metal / oneAPI / Vulkan backends plug into the same actor surface.
Agentic / LLM GPU actor blueprints on atomr-accel-cuda: RagPipeline, EmbeddingCache, CpuVectorIndex, SharedGpuStateCoordinator, LangGraphGpuActor.
CUB-backed device-wide reductions, scans, sorts, histograms, and selects, surfaced as an atomr actor compiled per-(op, dtype) via NVRTC against the atomr-accel-cuda Phase 0.6 disk cache.
GPU acceleration via the actor model. Wraps NVIDIA CUDA libraries (cuBLAS, cuDNN, cuFFT, cuRAND, cuSOLVER, cuSPARSE, cuTENSOR, cuBLASLt, NVRTC, NCCL) as supervised atomr actors with generation-validated buffers and a uniform async surface.
Interactive-rate GPU actor blueprints on atomr-accel-cuda: ImageFilterPipeline, ParticleSystemActor, SpatialIndexActor, ClothSimulationActor, FluidSimulationActor, plus NVRTC kernel sources.
CUTLASS kernel-template instantiation via NVRTC for atomr-accel. Provides GEMM, grouped GEMM, implicit-GEMM convolution, and EVT (epilogue visitor tree) actors that JIT CUTLASS C++ templates against the per-arch toolchain pinned by atomr-accel-cuda's NvrtcActor.
FlashAttention v2 + v3 kernel templates for atomr-accel — fp16/bf16/fp8, causal, varlen, ALiBi, sliding window, sink tokens, MQA/GQA, paged KV-cache, and chunked prefill, dispatched through NVRTC + Phase 0.6 cubin cache.
Universal GPU actor blueprints for atomr-accel-cuda: DynamicBatchingServer, InferenceCascade, ModelReplicaPool, FairShareScheduler, ModelHotSwapServer, SpeculativeDecoder, MoeRouter.
GPU observability for atomr-accel: NVTX ranges, NVML metrics actor, and CUPTI activity tracing. Implements the KernelTrace hooks exported by atomr-accel-cuda and registers atomr-telemetry probes.
TensorRT engine builder + runtime as supervised atomr actors. Wraps libnvinfer / libnvonnxparser at runtime (proprietary library, not vendored). ONNX import, INT8/FP8 PTQ calibration, IPluginV3 Rust trampolines, dynamic shapes, refit.
Distributed training blueprints on atomr-accel-cuda: DataParallelTrainer, PipelineParallelTrainer, TensorParallelTrainer, AsyncParameterServer, optimizer + loss enums.
Composable agentic framework on top of atomr actors and atomr-infer.
Agent actor + per-turn pipeline + tool-call orchestration loop.
Avatar domain types for real-time, embodied agents: ARKit blendshape weights, viseme→blendshape mapping, AvatarFrame timecoding, AvatarSink trait, and the Live Link CBOR wire format consumed by an Unreal Engine 5 MetaHuman receiver.
Avatar orchestrator: composes STT (perception), atomr-infer (cognition), TTS (synthesis), viseme→blendshape mapping, and the Sync-Manager into a supervised actor pipeline. Drives an AvatarSink (Live Link UDP to UE5 MetaHuman) with timecode-aligned audio + ARKit blendshape frames.
Avatar audio→ARKit-blendshape provider backed by NVIDIA Audio2Face-3D. Currently a documented stub: the real runtime lives upstream in atomr-infer once FR-A2F-001 (audio2face modality) lands. x86_64 + NVIDIA-GPU-only.
Avatar UDP Live Link sink: streams timecode-aligned audio + ARKit blendshape frames over UDP using the avatar-core CBOR wire format, ready for consumption by an Unreal Engine 5 ILiveLinkSource receiver plugin. x86_64-only (matches the UE5 tooling pipeline).
LLM cache (in-memory, semantic) + middleware integration.
Callable trait unifying agents, workflows, tools, and harnesses.
Channel + thread domain types for messaging-style transports (WhatsApp, Signal, Discord, in-memory). Defines the ChannelProvider trait, ChannelStore, ChannelEvent stream, ThreadTarget enum, HarnessInputAdapter, and the in-memory provider used for tests.
Channel orchestrator: attaches provider transports, supervises their lifecycle, routes inbound messages to bound Callable targets, queues outbound, persists state, and broadcasts ChannelEvent.
Axum web companion for the channel harness: per-provider webhook receivers (signature-verified), REST CRUD for channels/threads/messages, and a ChannelEvent /ws bridge.
Discord provider for atomr-agents channels. Gateway WebSocket (default) or Interactions webhook mode with Ed25519 signature verification.
Signal provider for atomr-agents channels via signal-cli JSON-RPC bridge (Unix socket or TCP).
WhatsApp Business Cloud API provider for atomr-agents channels. Outbound REST, HMAC-SHA256 webhook verification, and two-step media download.
atomr-agents CLI: harness eval run / registry list / harness pin / eval gate.
Uniform input/output contract and vendor trait for the coding-cli harness: request, result, normalized event schema, concept projection, and the CliVendor seam.