Tag

RealtimeSTT

Converts microphone or streamed audio to text with sub-second latency, pairing WebRTC/Silero voice-activity detection and wake-word activation with swappable local backends — faster-whisper by default, plus whisper.cpp, Moonshine, and sherpa-onnx.

Honcho

Provides a memory-first library and managed service that stores, reasons about, and serves long-term state for agents and users — offering continual representations, session context, vector search, and a chat-style API for personalized behavior.

Semantic Router

Routes LLM and agent decisions through semantic similarity instead of waiting for full generations, useful for intent routing, tool selection, guardrails, and multimodal handling.

Genesis World

Provides a scalable physics-and-rendering simulation interface for robotics and embodied-AI research — unified multi-physics solvers, the Nyx renderer, and the Quadrants compiler. Runs from laptop to datacenter GPUs; suited for sensor-rich data generation and RL/robotics prototyping.

pyannote/speaker-diarization-3.1

Performs speaker diarization (who spoke when) with pyannote-audio: combines voice-activity detection, speaker-change and overlapped-speech detection to produce time-stamped speaker segments; compatible with Hugging Face Endpoints and ASR pipelines.

torchtitan

Provides a PyTorch-native platform for experimenting with and scaling generative AI training, including composable parallelism, checkpointing, float8, logging, and Llama recipes.

Flash Linear Attention

Triton kernels and PyTorch layers for linear-attention, state-space, and sparse-attention token mixers (GLA, RWKV, Mamba2, GSA) as drop-in replacements for multihead attention. Runs on NVIDIA, AMD, and Intel GPUs with Hugging Face support.

LeRobot

A PyTorch-native, hardware-agnostic stack for robot learning: data collection, training, and deployment across 11+ robots, from SO100 to Unitree G1. Includes imitation, RL, and vision-language-action policies (ACT, Diffusion, Pi0, SmolVLA).

FlashInfer

GPU kernel library for LLM inference attention, sampling, and KV-cache, built on block-sparse formats with JIT-compiled customizable templates. Reports 29-69% inter-token-latency cuts vs compiler backends; powers SGLang, vLLM, and MLC-Engine.

Gemma

JAX-native library for loading, sampling, and fine-tuning DeepMind's open-weight Gemma LLM family, with multi-modal sampling and checkpoint utilities — aimed at researchers and engineers running experiments on CPU/GPU/TPU.

NVIDIA Earth2Studio

Chains pre-trained AI weather and climate models like GraphCast, Pangu, and FourCastNet into composable inference pipelines. Swap prognostic or diagnostic components, plug in reanalysis sources, and add ensemble perturbations or in-loop metrics.

Ant Design X

React components for building LLM chat and agent interfaces: message bubbles, prompt sets, conversation lists, and sender inputs under a RICH interaction paradigm, plus a streaming Markdown renderer and hooks for wiring UI to model data streams.