Fulling is a foundation for building dedicated AI workspaces that combine runtime, memory, and skills within a secure Kubernetes-based credential boundary.

A cross-IDE repository knowledge engine that creates a grounded Q&A system for codebases, allowing AI agents to provide accurate answers with file paths and line numbers.

An OpenAI-compatible serverless worker for RunPod that enables fast LLM deployment using the vLLM inference engine.

An open-source RAG backend platform that combines embedded SQLite and LanceDB databases with a spreadsheet-like UI to simplify the creation of AI-enhanced tables and chatbots.

A unified AI model gateway and agent coordinator that simplifies provider management, request routing, and remote control of AI agents.

A drop-in prompt compression library that reduces LLM input-token costs by 10-40% using loss-tolerance routing to protect critical code and tool calls while compressing prose.

Orthrus is a dual-architecture framework that enables high-speed parallel token generation for LLMs while maintaining the exact generation fidelity of the original base model.

A GenAI configuration framework that uses Jsonnet to manage prompts and chains as declarative configuration, enabling versioning and automatic hardware parallelism via WebAssembly.

Unbody is an open-source backend infrastructure project designed to provide a comprehensive data management and indexing system for AI applications.

An AI-powered penetration testing automation framework that orchestrates security tools and multi-agent AI to perform adaptive security assessments and evidence-backed reporting.

A transparent network interceptor for LLM tooling that allows for the inspection, transformation, and cross-provider routing of AI API traffic.

A Go kernel for Jupyter and nteract that allows users to run Go code interactively in a browser-based notebook for data science and documentation.

An API conversion tool that provides Gymnasium and PettingZoo bindings for popular external reinforcement learning environments like OpenAI Gym and DeepMind Control.

A management platform and optimized user interface for Dify applications, providing a professional front-end for Chatflow and Workflow AI agents.

A 2D agentic IDE that replaces terminal tabs and SSH with a zoomable infinite canvas for managing multiple machines, terminals, and AI agents in one unified space.

An MCP server that enables AI agents to perform binary analysis and reverse engineering using the radare2 framework.

A lightweight Android VPN proxy tool based on tun2socks that supports per-app proxying and remote control via the Model Context Protocol (MCP).

A lightweight macOS menubar app that tracks Claude Code usage limits and provides notifications when AI agent sessions need user attention.

A Claude Code plugin that turns AI-native development into a structured system with context-budgeted sprints, multi-agent orchestration, and automated quality verification.

A lightweight C++20 interface for GPU compute via WebGPU that simplifies shader dispatch and data movement across native and browser environments.

FuxiCTR is an open‑source Python library (PyPI package) for click‑through‑rate prediction. It bundles more than 40 modern CTR models—both classic factorization‑machine style and recent sequence‑aware architectures—implemented for PyTorch and TensorFlow. The library emphasizes configurability, automatic hyper‑parameter tuning, reproducible benchmarks, and easy extensibility, making it useful for researchers and engineers working on advertising, recommendation, or sponsored‑search systems.

A framework for generating knowledge-driven synthetic data using knowledge graphs to fill knowledge gaps in LLMs for improved supervised fine-tuning.

A local-first document anonymization workbench that uses semantic NER and visual grounding to redact sensitive text and visual elements from unstructured files.

A set of APIs that allow developers to easily move TensorFlow and Keras model training from local environments to distributed training and tuning on Google Cloud Platform.

A curated registry and CLI for discovering, installing, and managing human-authored skills for AI agents across multiple platforms.

A Windows utility that converts images to a 40-color palette and automatically fills 24x24 pixel art canvases in Arknights via computer vision and mouse automation.

A Python library that provides easy integration between pandas DataFrames and various AWS services, simplifying data movement and management for data lakes and databases.

A framework that turns Python notebooks into interactive web applications, allowing users to deploy chats, AI agents, and dashboards directly from .ipynb files.

A C++ programming model and core libraries that enable performance portable applications to run efficiently across all major HPC platforms.

torchtune is a PyTorch‑based library that provides ready‑to‑run recipes and configs for fine‑tuning, RL‑HF, distillation, quantization‑aware training, and inference of many modern LLMs (Llama, Gemma, Qwen, Phi, etc.). It abstracts distributed training, memory‑saving tricks, and device support behind a simple CLI (`tune`) and YAML files, integrating with Hugging Face, torch‑ao, FSDP2, and logging tools. Though development stopped in 2025, it remains a practical, open‑source toolkit for LLM experimentation.

An AI-native ecommerce search platform that uses semantic search and personalization to improve product discovery and conversion for online brands.

A comprehensive scientific computing system for OCaml that provides n-dimensional arrays, linear algebra, and deep learning capabilities for high-performance analytical code.

An embeddable AI agent harness kernel written in Rust that provides the execution loop, memory, and coordination infrastructure for agent-powered applications.

MushroomRL is a modular Python library for reinforcement learning that provides a wide array of classical and deep RL algorithms and integrates with popular tensor libraries and benchmarks.

hf‑mem is a lightweight Python CLI (and library) that estimates the RAM required to run any Hugging Face model that uses Safetensors or GGUF weights. It fetches only metadata via HTTP range requests, supports Transformers, Diffusers, Sentence‑Transformers, and can optionally estimate KV‑cache memory and MoE breakdowns. Installable as a standalone tool, a Hugging Face CLI extension, or an agent skill.

A framework for efficient, globally distributed AI model training over the internet, featuring fault-tolerant communication and optimized network bandwidth utilization.

A JAX-native library for Multi-Agent Reinforcement Learning that provides GPU-accelerated environments and baseline algorithms for efficient benchmarking.

A lean library for training and enable loading k-sparse autoencoders (SAEs) and transcoders on HuggingFace language model activations to aid in mechanistic interpretability.

A private enterprise control plane for AI agents that unifies model routing, skill distribution, and knowledge management across different local agent runtimes.

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

A local, database-first code intelligence and SAST platform that turns codebases into queryable facts to reduce token costs for AI agents and security teams.

A checkpointing and persistence library for JAX models that enables efficient saving and restoring of model states, including support for asynchronous checkpointing.

vLLM‑Moet is a patched vLLM fork that adds 2‑bit expert quantisation, FP4 recovery, tiered GPU/host/NVMe expert storage, and speculative decoding to run massive MoE models (e.g., 753 B GLM‑5.2) on consumer‑grade RTX PRO 6000 or RTX 5090 GPUs. It ships Docker images, custom SM 120 kernels, and a set of configurable knobs for memory‑speed‑quality trade‑offs.

A comprehensive benchmark for Physics-Informed Neural Networks (PINNs) that implements multiple variants and a challenging dataset to evaluate their performance in solving partial differential equations.

A native Rust client for Qdrant, enabling high-performance vector search and database management within Rust applications.

A high-performance automatic differentiation package for Julia that supports mutation and leverages optimized intermediate representations for efficient derivative calculations.

MDia is a Python library that formalises LLM intermediate reasoning as reusable “dialect cards” (LSFs) and provides a deterministic eight‑stage pipeline (collect → create → evolve → profile → select → run → validate‑rules → report) for generating, evolving, profiling, routing, and auditing these protocols across heterogeneous models. It supports black‑box LLM APIs, offers several routing plans, includes a 100‑rule bank, and guarantees reproducibility by freezing decisions before test evaluation. The repo includes a toy offline demo that runs without any API keys, extensive documentation, MIT licensing, and a citation to the ICML 2026 poster and arXiv paper.

An AI-powered code review tool that uses DeepSeek models to automate PR reviews via GitHub Actions or local CLI audits.