img2threejs is a Python‑backed skill that lets an LLM turn a single reference image into a procedural Three.js model. It builds a spec, runs a gated vision‑review loop, and emits TypeScript code that recreates the object from primitives and shaders, ready for animation. The system is extensible via domain plugins (e.g., CS2 weapon skins) and includes a live demo gallery.

An AI agent-led search engine that aggregates and scores real-time human engagement from social platforms and technical forums to provide current, high-signal briefs.

gpt‑instruct is a version‑controlled prompt‑and‑evaluation toolkit for OpenAI’s Codex model. It ships stable (gpt‑5.6‑sol‑v45) and early‑pre‑release (gpt‑6‑astra‑v1‑rc1) prompt packs, a strict A‑B‑C testing gate, and a Python CLI (`codex‑instruct.py`) that can apply, snapshot, and roll back instruction files safely.

A high-accuracy document parsing engine that converts PDF, DOCX, PPTX, XLSX, and images into structured Markdown or JSON for LLM and RAG workflows.

An open‑source AI‑agent skill that structures math‑modeling competition work into three stages (analysis, coding, paper), supports Python/MATLAB, generates publication‑grade figures, enforces reproducibility, and outputs Word/LaTeX papers with built‑in quality‑check subagents. Includes a DeepSeek Harness plugin and a full test suite.

A tool for generating and verifying DOCX documents that strictly follow the GB/T 9704-2012 national standard for official Chinese government documents.

A local inference router that distributes AI workloads across multiple compatible computers on the same network using OpenAI and Ollama compatible endpoints.

An agent platform for deploying secure AI coworkers with isolated browser/shell environments and a strict governance gateway for auditing and policy enforcement.

An experimental Vulkan 1.4 implementation that replaces traditional graphics API abstractions with GPU pointers and application-owned descriptor heaps to simplify GPU programming.

An AI-powered real-time global intelligence dashboard for news aggregation, geopolitical monitoring, and infrastructure tracking.

An open-source RAG engine that combines deep document understanding and agentic capabilities to create a high-fidelity context layer for LLMs using complex unstructured data.

An AI agent skill for generating professional single-file HTML presentations with built-in presenter mode and two distinct high-end visual styles.

A Memory Agent that manages the long-term memory of AI agent fleets by curating, reconciling, and expiring information to prevent data noise and vendor lock-in.

A macOS application for discovering, organizing, and editing AI coding agent skills and agents across multiple tools like Claude Code, Cursor, and Windsurf.

An open-source, agentic-first CRM where AI agents autonomously research leads, record factual evidence, and manage follow-ups instead of relying on manual human entry.

A data visualization skill for AI agents that generates editorially-styled charts and full-page HTML reports using specific visual grammars for different reading speeds.

AirLLM enables the execution of massive LLMs on low-VRAM consumer GPUs by streaming model layers one at a time instead of loading the entire model into memory.

A bilingual, step‑by‑step tutorial that teaches LLM inference fundamentals and guides you to build a minimal SGLang‑style inference engine, then deep‑dives into the real SGLang code and how to contribute to it.

PaddleOCR is an open‑source OCR and document‑AI toolkit that converts images/PDFs into structured, LLM‑ready JSON or Markdown. It offers multilingual scene‑text models (PP‑OCRv6), a lightweight vision‑language parser (PaddleOCR‑VL‑1.6) for tables, formulas, and ancient scripts, and a structure‑aware pipeline (PP‑StructureV3). Models run on CPU, GPU, XPU, NPU or via ONNX/TensorRT, and integrate with RAG/agent platforms like Dify and RAGFlow.

An open-source AI gateway and Python SDK that provides a unified OpenAI-compatible interface for calling 100+ different LLM providers.

UX/UI Agent Skills is a Claude‑agent knowledge kit that provides design tokens, component specs, accessibility audits, and 17 slash‑command‑style “skills” to let the AI act as a senior UI/UX architect. It can generate token‑driven code for any framework (React, Vue, SwiftUI, Flutter, etc.), enforce WCAG 2.2 compliance via 37 CI gates, and includes a library of 138 design systems for aesthetic direction. Install via `npx ux-ui-agent-skills init`; the agent reads a `CLAUDE.md` brief and runs the skills or natural‑language prompts to produce design assets, code, and audits.

VODER is a local, offline, professional-grade voice processing toolkit that combines 8 modes—speech-to-text, text-to-speech, voice conversion, music generation, speech enhancement, sound effects, vocal separation, and speaker diarization—plus a Project Eva DLC for image/video/3D generation and uncensored chat. It runs entirely on your machine, works with or without a GPU, and uses open-source models like Whisper, Qwen3-TTS, Seed-VC, and more.

A native Rust CLI that lets AI agents (or scripts) control Chrome via simple commands, using accessibility‑based element refs (`@eN`). It supports opening pages, taking snapshots, clicking/filling, extracting text, screenshots, network mocking, tab management, batch execution, and an experimental WebMCP tool system.

A local-first search layer that unifies ripgrep, BM25, and vector search to help humans and AI agents efficiently retrieve evidence from local workspaces.

dsh‑web is a plug‑in aggregation package that extends DeepSeek Harness’s web UI with a full AI‑development workbench: task board with cron, mobile remote control, SSH terminal, image‑understanding tool, per‑model capability UI, Git graph, skins, pets, rescue mode, and more. Install via `dsh plugin --profile web add @linxin666/dsh-web-all@latest` and restart DSH.

A whole-brain leaky integrate-and-fire emulation of the adult fruit fly based on the FlyWire connectome, allowing researchers to simulate and analyze neural spike propagation.

A framework for building agents and LLM-powered applications that simplifies development through interoperable components and a vast library of third-party integrations.

An automated ID photo processing tool that uses AI to handle face detection, background replacement, and layout arrangement for professional-grade ID photos.

Mem0 is a memory layer for AI agents and assistants that enables personalized interactions by storing and retrieving user preferences and history over time.

A self-hosted AI chat platform that unifies all major AI providers into a single, privacy-focused interface with support for agents, code execution, and multi-user management.

An AI prompt optimization tool that iteratively improves prompts for text and image generation models to enhance output quality and accuracy.

An unrestricted, open-source AI studio for generating images, videos, and audio using 400+ models without content filters or subscription fees.

A development harness for the Pi coding agent that enforces Spec-Driven Development (SDD), strict TDD evidence, and subagent orchestration to turn AI coding into a disciplined engineering process.

A persistent memory compression system for Claude Code that preserves project context across sessions using semantic summaries and hybrid search.

An MCP server that enables AI agents to automate native iOS and Android apps across simulators, emulators, and real devices using a platform-agnostic interface.

A model-definition framework providing a unified API to access, train, and deploy state-of-the-art pretrained models across text, vision, audio, and multimodal domains.

An automated generator for Cloudflare WARP MASQUE configurations that runs on GitHub Actions or Cloudflare Workers, supporting proxy chaining to change exit countries.

A low-level orchestration framework for building, managing, and deploying long-running, stateful AI agents with durable execution and human-in-the-loop capabilities.

DaVinci Resolve MCP is an open‑source server that exposes the full DaVinci Resolve Studio scripting API as a JSON‑over‑HTTP service. AI agents can call high‑level tools (project creation, media import, timeline editing, grading, Fusion, Fairlight, render setup, etc.) and receive a uniform result envelope for safe, auditable automation. It works with the paid Studio edition via the normal external‑scripting interface, and includes a Lua‑based bridge to reach the free edition. An optional Node‑based “advanced” server can edit Resolve project files offline. The package ships a local browser control panel and integrates with many IDEs, making it a practical foundation for on‑premise AI post‑production assistants.

A visual registration and configuration tool for Cloudflare WARP / MASQUE that simplifies account setup and generates compatible configurations for multiple proxy clients.

VibeVoice is a family of open-source voice AI models providing long-form speech recognition (ASR) and text-to-speech (TTS) capabilities with high speaker consistency and efficiency.

A high-performance C++ audio inference framework built on ggml that provides a portable, native runtime for running a wide variety of local audio AI models.

The repository is the production‑grade code for X’s “For‑You” feed. It merges in‑network (followed) and out‑of‑network (retrieved) tweets, scores them with a transformer model (Phoenix), applies configurable weights, runs visibility‑filtering based on a rich label system, and finally blends in ads and suggestions. The repo includes the request‑time pipeline (home‑mixer), retrieval models (phoenix, simclusters), labeling infrastructure (agatha, clip, botmaker, scarecrow), and a transparency tool. Recent updates add weight documentation, a Brazil‑election filter, and synthetic training code. Licensed under Apache 2.0.

An MCP server that provides AI coding agents with IDE-level semantic tools for symbol-aware retrieval, editing, and refactoring across large codebases.

A framework for forward deployed engineers to manage client briefs, decisions, and delivery evidence for AI projects using AI coding agents and local Markdown records.

Handy is an open‑source, cross‑platform desktop app that does offline speech‑to‑text using Whisper or Parakeet models. You press a global shortcut, speak, and the transcribed text is pasted into any window. Built with Tauri (Rust + React), it runs on Windows, macOS, and Linux, supports GPU acceleration, and offers CLI flags and Raycast integration for power‑user control.

A tool for distilling the cognitive frameworks and mental models of public figures into runnable AI skills for agent runtimes.

A coding-agent skill that enables the creation of stunning, single-file HTML presentations from scratch or by converting PowerPoint files, using a visual style discovery process.