Trending AI projects

A unified real-time search engine skill for AI agents that supports general web search, vertical domain search, and full-page content extraction.

CPA Manager Plus is a self‑hosted web panel that monitors and manages the CLIProxyAPI (CPA) gateway for large‑language‑model providers. It stores request logs in SQLite, visualises cost/usage, diagnoses failures, tracks quota health, and offers both a lightweight UI replacement and a full‑mode manager server. Deployable via Docker or native packages, it keeps all data on‑premise with encrypted keys and no telemetry.

An automated pipeline that transforms text or images into interactive 3D holographic cards with depth and shimmer, delivering both a web viewer and an editable Blender project.

A deployment toolkit for running the Qwen3.8 Flash Next model on a single NVIDIA DGX Spark, providing an OpenAI-compatible API for text and vision tasks.

WeThinkIn/AIGC-Interview-Book is a free, open-source Chinese-language platform for AI job seekers — covering interview prep, technical knowledge (AIGC, LLMs, AI agents, deep learning, etc.), study routes, career planning, and a community contribution system. It emphasizes learning only technologies with proven long-term value, and is maintained by a lead editor plus six co-editors, all with industry experience.

A curated, open-source list of AI engineering interview questions and answers, organized into 15 topic sections, maintained by Amit Shekhar of Outcome School, with many answers linking to external blog posts and videos.

这是一个深度学习笔记合集，涵盖CV、NLP、大模型和Agent，配套视频教程、数据集、交流群及就业指导服务。

An AI-assisted PCB design plugin for KiCAD 10 that allows AI assistants to directly create schematics, lay out PCBs, and run design audits using the Model Context Protocol.

NVIDIA Run:ai is an open-source platform for orchestrating and managing GPU resources on Kubernetes, enabling efficient sharing, scheduling, and governance of AI workloads across teams.

A Python SDK for the Model Context Protocol that enables developers to build standardized servers and clients to connect LLM applications with external data and tools.

EchoMuse transforms second-hand Amazon Echo Dot 2nd gen devices into private, local voice satellites for Home Assistant, removing all cloud dependencies and Amazon services.

An open-source agent microharness that enables persistent, self-guided AI agents with a continuous inner monologue and Bash-based execution.

A high-speed, local decision-making engine that provides yes/no, multiple-choice, and scoring answers to text-based questions on a laptop CPU.

bili-cli is a tool that empowers AI agents to manage Bilibili accounts, enabling them to handle fan interactions, video uploads, content analysis, and channel analytics automatically.

dsh-synapse is a plugin for DeepSeek Harness that visualizes conversations, follow-up questions, and branches as a browsable, draggable, zoomable map, while keeping DSH's native sessions as the single source of truth. It stores only layout and short previews, reads message content on demand, and never alters prompts or model behavior.

Surf is a command-line tool that lets AI agents control Chrome via simple commands and a browser extension, offering zero config, agent-agnostic use, and the ability to query AI services without API keys.

A terminal-based pixel-art dashboard that visualizes multiple AI coding agents as coworkers in a virtual office to monitor their status and token usage.

A web-based management system for Xianyu that automates customer messaging, AI-powered replies, and product order fulfillment across multiple accounts.

A high-performance CLI tool and visualization dashboard for tracking token usage and costs across multiple AI coding agents.

fast-jev-compaction is an npm library and Claude Code plug‑in that replaces Claude Code’s default summarising compaction with a selective pruning process. It sends the full conversation (with tool results replaced by short placeholders) to the Jev LLM, asks whether each tool call and its result should be kept, and drops only those deemed unnecessary. The result is a transcript where all user/assistant text stays verbatim and only unneeded tool interactions are removed.

A simulation framework for evaluating conversational AI agents on complex, real-world customer service tasks across text and full-duplex voice modalities.

The repository is a genuine, AI‑related project: a curated collection of Claude‑compatible plugins, skills and prompt templates for German legal work.

Apache Airflow is a code-based platform for authoring, scheduling, and monitoring workflows (DAGs). It's best for mostly static pipelines, supports ML/AI orchestration, and emphasizes idempotent tasks with no large data passing. It runs on POSIX systems, requires specific Python/database versions, and is installed via pip with constraint files. The project follows strict SemVer and is governed by the Apache Software Foundation.

A local-first, open-source AI agent that runs a control loop on your machine, can operate a browser, edit files, run approved commands, and keeps persistent memory — designed to work efficiently with small quantized models via llama.cpp.

A high-performance LLM inference engine optimized for agentic workloads that separates its C++ control plane from its Python execution plane for maximum speed and flexibility.

FluidAudio is a Swift SDK that brings fully local, low‑latency speech AI—ASR, TTS, VAD, diarization and enhancement—to iOS and macOS by running optimized CoreML models on the Apple Neural Engine.

Qwen‑Image‑2.1 is a 7 B‑parameter diffusion model for text‑to‑image generation and image editing (including transparent outputs). It offers a fast 8‑step “Turbo” variant, supports up to 10 reference images, and integrates with Diffusers, ComfyUI, vLLM‑Omni, SGLang, LightX2V, and OpenVINO. Hosted Pro/Turbo APIs are available on Alibaba Cloud.

A lightweight agentic RL framework that allows developers to train AI agents with reinforcement learning using their existing harnesses and environments with minimal code changes.

A Python package for generating text and fine-tuning large language models specifically optimized for Apple silicon using the MLX framework.

Omalauncher is an Android launcher that mimics Omarchy’s look and can run an Arch Linux ARM userland via proot, letting each Linux app appear as its own Android task. It supports themed UI, editable OWL widgets, a bottom bar that forwards prompts to AI coding agents, and works without root, ads, or accounts.

A governance standard and toolset for knowledge repositories maintained by LLM agents, providing structured rules, resumable task state, and verifiable evidence of work.

A standalone extension for the Pi coding agent that reduces token costs and inference work by optimizing tool calls, observation handling, and context compaction.

An efficient portrait animation tool that animates static images of humans and animals using driving videos or images with precise stitching and retargeting control.

An idiomatic Java library that simplifies integrating LLMs into JVM applications through unified APIs for providers and embedding stores, and a toolbox for building RAG and agentic apps.

MotionClone is a Windows‑only app that uses your ChatGPT/Codex login to analyse a short motion‑graphics video and rebuild it as an editable HyperFrames (HTML‑based) project, offering side‑by‑side comparison and export to MP4 or source code.

A framework that bundles LLMs and speech-to-text tools into single-file executables for zero-installation, cross-platform local execution.

A plugin for the DeepSeek Harness desktop app that enables remote control and interaction with AI agents via WeChat, supporting file transfers, voice commands, and session management.

A Triton-based domain-specific language that simplifies writing high-performance GPU kernels using tensor-oriented meta-programming.

A Model Context Protocol (MCP) server and remote script that connects Claude AI to Ableton Live for prompt-assisted music production and track manipulation.

An AI-powered executive management system that uses a council of specialist agents to handle strategic and operational leadership tasks while acting as a digital twin for leaders.

LiteParse is a fast, local, open-source PDF parsing tool that extracts spatial text and structure into Markdown or JSON without cloud dependencies.

CapCut Mate API is an open-source automation tool that provides a RESTful interface for Jianying (CapCut) draft editing, enabling developers to programmatically create, edit, and render videos.

A software wrapper that enables DLSS-style neural rendering, frame generation, and ray reconstruction on AMD GPUs for Direct3D 12 games.

fallow is a static analysis tool for TypeScript and JavaScript codebases that builds a repository graph to detect complexity hotspots, duplication, unused code, architecture drift, and styling issues. It runs as a single Rust binary with no TypeScript compiler or configuration needed, and can gate pull requests on new findings.

JourniOne‑Planning‑Skills is an open‑source Node.js skill that adds AI‑driven travel planning to the JourniOne platform. It lets users describe a trip in natural language, generates day‑by‑day itineraries, searches hotels (via TourMind) and flights (via Kiwi.com), compares prices and cancellation terms, and produces a visual, shareable travel journal. Installation requires the TourMind hotel skill and Kiwi MCP; the backend LLM lives on journione.ai.

kicad‑happy is a Python‑only, AI‑enhanced KiCad design‑review tool. It parses schematics/PCBs, builds a power‑tree, runs rule‑based checks (ESD, decoupling, trace widths), generates SPICE test‑benches, evaluates EMC compliance, cross‑references datasheets, and creates a full BOM with supplier queries. It can be used locally, as a GitHub Action, or as plug‑ins for Claude Code, OpenAI Codex, Google Antigravity, and opencode, enabling natural‑language queries like “Check my boost converter for EMI”.

A professional financial skill set for Claude adapted for the Chinese A-share market, integrating Wind, iFind, and AkShare for automated financial analysis and reporting.

A multi-agent offensive-security framework that turns existing AI coding agents into automated zero-day hunters for authorized security testing.