TenSEAL is a Python library for homomorphic encryption on tensors, built on Microsoft SEAL. It lets you run arithmetic operations on encrypted data, enabling privacy-preserving machine learning without exposing raw information.

A framework-free, self-evolving AI agent system built in pure Python that provides 24/7 personal assistance with multi-tenant routing and a three-layer memory system.

An agent testing framework that uses simulations, user simulators, and judge agents to evaluate the behavior and robustness of AI agents in multi-turn conversations.

An enterprise AI resource management platform that helps organizations control AI costs, manage model access, and quantify AI productivity through centralized identity and budget tracking.

A modular runtime and orchestration system for AI agents that automates software development lifecycles through structured pipelines, quality gates, and specialized agent roles.

A framework for packaging, running, and tracking machine learning experiments across local environments, Kubernetes, and Google Cloud Platform.

A C++17 library for simplifying CUDA kernel benchmarking, offering parameter sweeps and detailed throughput and bandwidth measurements.

Go bindings for llama.cpp that enable high-performance local LLM inference within Go applications using the GGUF format.

Nengo is a Python library for building and simulating large-scale neural models, supporting both spiking and non-spiking simulations.

A Python implementation of TypeAgent KnowPro for building structured RAG systems, serving as an experimental prototype for structured retrieval.

A library for simplifying the fine-tuning, inference, and retrieval of ColBERT late interaction models, providing high-speed indexing and flexible training options.

An open-source inference server that provides a standardized REST and gRPC interface for serving machine learning models across various frameworks.

A terminal-native coding agent built with TypeScript and Node.js that integrates LLMs with local file and shell tools for automated software development.

A C++ library and file format for compressed 3D Gaussian splats that reduces file sizes by up to 10x compared to .ply files with minimal visual difference.

A cross-platform AI assistant app for Android, iOS, and macOS that enables real-time chat, multimodal analysis, and the instant creation of web applications using various LLM providers.

A diverse suite of scalable, hardware-accelerated reinforcement learning environments written in JAX for high-speed RL research and experimentation.

An enterprise-grade secure gateway that provides auditing, governance, and access control for AI API calls and MCP tool invocations across an organization.

An open-source engine for fine-tuning and serving large language models, providing tools to deploy models on hosted or self-managed Kubernetes infrastructure.

A KV-cache compression method for long-reasoning LLMs that uses trigonometric frequency-domain compression to reduce memory usage by 10.7x and boost throughput by 2.5x without accuracy loss.

A local-first desktop AI workbench that unifies multiple LLM providers, AI agents, knowledge bases, and MCP tools into a single, privacy-focused interface.

An Android SDK for facial recognition and 3D passive liveness detection to prevent spoofing in security and fintech applications.

An MCP server that enables LLMs to manage Docker containers, images, and volumes using natural language commands.

A fully-local session memory tool for Claude Code and OpenCode that automatically logs activity and creates compact summaries to eliminate the cold-start problem.

A training and modeling framework for SpeechLM-based text-to-speech systems, supporting pre-training, SFT, and RLHF alignment across single or multi-GPU clusters.

A provider-agnostic CLI AI coding assistant that lets developers use any LLM provider, including local models and existing subscriptions, to edit codebases and automate workflows.

A PyTorch-native agentic RL library that separates infrastructure concerns from model concerns to simplify scalable reinforcement learning experimentation.

OpenYak is an open‑source desktop GUI (Electron + React) that lets you chat with Codex or Claude Code agents while keeping all conversation history, files and project data locally (Rust + SQLite). It integrates the official runtimes via stdio, shows referenced files (Markdown, HTML, PDF, DOCX, code), and offers a shared Playwright‑driven browser where both user and agent can control the same page. Run it with Node 26, Rust 1.90, and your own Codex/Claude CLI binaries; the app is in v2 alpha and licensed Apache‑2.0.

A local workbench that turns documents into an inspectable 3D knowledge graph by wrapping Microsoft GraphRAG with a visual interface for indexing and exploration.

An open-source e-commerce system based on Laravel 13 that integrates AI models and Model Context Protocol (MCP) for intelligent automation.

Skore is a Python library that structures machine learning experiments by reducing evaluation boilerplate and providing built-in methodological guidance for data scientists.

An interactive web application that uses Groq's LLM API to generate and modify web applications based on natural language queries.

A fast and extensible library for training, evaluating, interpreting, and serving decision forest models like Random Forest and Gradient Boosted Trees.

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.

A framework and benchmark for developing Multimodal Large Language Models capable of deep research through iterative visual and textual search across images and videos.

A private, browser-native AI studio and agent workspace that runs LLMs locally on your device using WebGPU, requiring no API keys or backend servers.

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.

A local AI CLI unified gateway that provides a single entry point for tools like Claude Code and Gemini CLI, featuring intelligent failover, usage tracking, and workspace isolation.

An AI-powered tool that keeps multilingual GitHub documentation current by automating the translation of Markdown, Jupyter notebooks, and images while preserving repository structure and links.

Custom ComfyUI nodes that implement Microsoft's TRELLIS.2 model to generate high-quality 3D meshes with PBR materials from a single image.

A framework for configuring and running large-scale machine learning experiments on top of Jax, optimized for Cloud TPU and GPU infrastructure.

An MCP server that enables AI language models to interact with YouTube by providing tools to search videos, retrieve transcripts, and analyze channel data.

A local Node.js server that lets AI‑assistant clients (via MCP) run SQL queries on a personal Apple Health CSV export, using an in‑memory DuckDB database. It provides schema discovery, ad‑hoc queries, and health‑summary reports while keeping data on the user’s machine.

A library providing a programming model for writing performance-critical machine learning kernels that are portable across different GPU and CPU architectures.

An open-source project management platform that uses conversational AI and browser automation to execute tasks and manage workflows via natural language.

KISS Sorcar is an open‑source, local‑first AI‑agent framework that runs a daemon (`kiss‑web`) and provides VS Code, web/mobile, and Python interfaces for issuing natural‑language tasks. It supports any LLM you configure (OpenAI, Anthropic, Gemini, local endpoints, etc.), mixes multiple models in one workflow, and includes 43 built‑in third‑party agents for messaging, email, smart‑home, and productivity services. Custom Python “extension agents” let you define prompts, tools, budgets, and safety hooks per task. All prompts stay on your machine; the system handles git worktree isolation, voice wake‑word, scheduled automations, and tool calling, making it a privacy‑focused, extensible personal AI assistant.

OSCAR is a 2-bit KV cache quantization method that uses offline spectral covariance-aware rotations to reduce memory usage by 8x while maintaining near-BF16 accuracy.