A lightweight, plug-and-play framework for identity-preserving video generation that maintains high subject consistency by training only 1% of additional parameters.

A unified framework for the inference, evaluation, and post-training of multimodal models, supporting image understanding, generation, and editing through a single interface.

A Python library that transcribes audio using faster-whisper and uses LLMs to translate and polish the text into synchronized .lrc subtitles.

A multi-view, multi-source benchmark and baseline for drone-based geo-localization, enabling the matching of drone, satellite, and street-view images to locate buildings.

An OpenRouter provider for the Vercel AI SDK that enables access to 300+ LLMs and embedding models through a unified interface.

Flow‑Next is a plugin for coding agents (Claude Code, OpenAI Codex, Cursor, etc.) that turns a change request into a durable spec, then automatically drives planning, implementation, adversarial review, live QA and PR creation, recording all evidence in the repository.

LLM‑Subtrans is a Python‑based desktop and CLI tool that translates subtitle files (SRT/ASS/VTT) using LLM APIs (Gemini, OpenAI, Anthropic, etc.). It supports pluggable subtitle formats, optional audio transcription, project‑file resume, and many advanced CLI options. Install via pre‑built Windows/macOS packages or from source with a virtual‑env installer; run through a simple GUI or command line, providing the provider’s API key.

An MCP server that connects AI assistants to Google Docs, Sheets, Drive, Gmail, and Calendar, enabling LLMs to read and manage Workspace data.

An MCP server providing LLMs with real-time and historical financial data for Borsa Istanbul, US stocks, TEFAS funds, and crypto, enabling AI-driven financial analysis.

A local LLM routing proxy that reduces API costs by 40-70% by routing prompts to the cheapest viable model and escalating to premium models only when necessary.

A minimal and scalable JAX-based research codebase designed for rapid LLM iteration and automated AI research conducted by both humans and AI agents.

TorchANI 2.0 is a PyTorch library for training and using ANI‑style neural‑network interatomic potentials, providing GPU‑accelerated energy/force predictions and tools for molecular dynamics and ML/MM simulations.

FlagScale is a unified toolkit for the FlagOS ecosystem that provides a single interface for training, reinforcement learning, and inference of large models across diverse AI hardware chips.

IgFold is a deep learning tool that predicts 3D antibody structures from amino acid sequences, enabling fast and accurate structural modeling for drug discovery.

RFUAV provides a publicly‑available RF‑signal dataset for UAV detection/identification, along with Python/MATLAB tools to turn raw IQ recordings into spectrograms, train vision‑style classifiers (ResNet, ViT, etc.) or YOLOv5 detectors, and evaluate performance across a wide SNR range.

An LLM-powered code completion package for Emacs that supports multiple providers and both chat-based and Fill-in-the-middle completion modes.

A personal AI agent for macOS that integrates voice, vision, and system access to manage meetings, calls, and digital tasks while autonomously improving its own code.

mosaic is a JAX‑based Python library that provides a unified, gradient‑optimizable interface for dozens of protein‑property and structure‑prediction models. By letting users compose differentiable loss terms (binding affinity, stability, solubility, inverse‑folding likelihood, etc.) and run continuous‑relaxation design with custom optimizers, it enables multi‑objective protein engineering in a single, JIT‑accelerated framework. The repo includes many state‑of‑the‑art models (Boltz, AlphaFold, OpenFold‑3, Protenix, ProteinMPNN, ESM, etc.), example notebooks, and detailed documentation, but it requires users to tune hyper‑parameters and have a GPU/TPU‑compatible JAX installation.

A ComfyUI wrapper for Tencent's Hunyuan3D-2.1 model that enables the generation of textured 3D models within a visual node-based workflow.

An AI-powered video enhancement toolkit specialized for anime that provides upscaling, motion interpolation, and restoration in a single processing pass.

A prompting framework that uses structural templates instead of concrete examples to improve LLM reasoning, providing a formal mathematical foundation for automated prompt engineering.

A library that extends the safetensors format to provide a unified storage and management system for various LLM compression and quantization schemes.

An MCP server that integrates Home Assistant with LLMs, allowing AI assistants to control devices, manage dashboards, and troubleshoot automations via natural language.

A web-based multi-voice TTS application for creating narrated stories and audiobooks, supporting a wide array of local and cloud-based speech engines.

usecomputer is a cross‑platform CLI (and TypeScript library) that lets AI agents take screenshots, map coordinates, and perform mouse/keyboard actions (click, drag, type, scroll) on macOS, Linux and Windows. It uses a native Zig backend, provides a coord‑map system for accurate screen‑to‑image mapping, streams global input events via SSE, and supports Kitty Graphics Protocol for agent‑friendly image delivery.

Sekai is a high-quality egocentric video dataset featuring over 5,000 hours of annotated first-person and drone footage for world exploration and generation.

A high-quality benchmark dataset and workflow for evaluating the physical understanding and realism of generative video models using real-world footage.

Vista4D is a video reshooting framework that uses 4D point clouds to synthesize dynamic scenes from source videos using novel camera trajectories and viewpoints.

EdgeCrafter provides compact Vision Transformers (ViTs) optimized for edge devices, enabling high-performance object detection, instance segmentation, and pose estimation with low latency.

CodexLoom is a self‑hosted platform that adds durable identities, profiles, and organizational tooling to Codex chat agents, enabling long‑running “Domain Agents” that can be accessed from multiple devices, communicate via bounded messages, and be exposed to external chat platforms (Feishu, Slack, Parall) through Interface Agents. Targeted at advanced solo users, it provides UI, CLI, and governance features, and is released under the Elastic License 2.0.

A research initiative focused on fundamental technologies to improve LLM prompting, inference speed, and context handling for building AI products.

A local-first data curation and annotation tool for computer vision datasets that uses embeddings to help users efficiently manage and label image and video data.

A library for training and evaluating deep neural networks that extract and match local visual features, supporting state-of-the-art models like LightGlue and GlueStick.

An open-source container platform that allows users to deploy and manage applications and AI services on Kubernetes via a graphical interface without requiring Kubernetes expertise.

A distributed caching platform that accelerates data-intensive analytics workloads by bridging the gap between computation frameworks and various storage systems.

A set of libraries for logging and visualizing multimodal robotics data via streaming or MCAP files, with support for Python, C++, Rust, and ROS.

A lightweight C++ toolkit providing a unified API to deploy over 300 AI models using backends like ONNXRuntime, MNN, and TensorRT.

WildClawBench is an end-to-end agent benchmark that tests AI agents' ability to perform complex, real-world tasks in live environments using real tools like browsers and bash terminals.

A lightweight CLI tool that reduces AI token costs by filtering CLI output and compressing file content before it reaches an AI agent.

A no-code GUI and framework for fine-tuning large language models, enabling users to experiment with hyperparameters and advanced training techniques without coding.

OpenVINO Model Server (OVMS) is a C++ inference server that serves LLMs, vision‑language models, and classic deep‑learning models via OpenAI‑compatible REST or KServe gRPC APIs. Optimized for Intel CPUs, GPUs and NPUs, it provides features like continuous batching, streaming, model versioning, hot‑reload, and Prometheus metrics. Deployable via Docker, bare‑metal binaries, or Kubernetes, it supports many model formats (TensorFlow, ONNX, OpenVINO IR, GGUF, etc.) and integrates with standard client libraries.

A low-level computer vision library written in Rust that provides fast, zero-copy image processing on CPU and NVIDIA GPUs for real-time ML pipelines.

A tool for converting ONNX models into LiteRT, TFLite, TensorFlow, and native PyTorch code to facilitate cross-framework deployment.

A Python library for time series forecasting that integrates scikit-learn compatible models, statistical methods, and foundation models into a unified API.

A high-performance automatic differentiation plugin for LLVM and MLIR that synthesizes fast gradients for optimized code, including GPU kernels and parallel programs.

A comprehensive collection of pre-converted Core ML models and sample apps for implementing on-device AI on iOS and macOS.

An open-source, self-hosted agentic harness for internal AI teams to build sovereign AI workflows with sandboxed code execution and model independence.

An agentic, local-first RAG system that provides citation-aware answers to questions about documents, featuring multimodal search and sandboxed Python analysis.