Trending AI projects

KTransformers is a research framework for efficient LLM inference and fine-tuning via CPU-GPU heterogeneous computing, enabling large MoE models like DeepSeek-R1 to run on a single consumer GPU (24GB VRAM) with 3-28x speedups, and offering fine-tuning with 6-12x speedup over ZeRO-Offload.

A bridge that connects QGIS to AI agents via the Model Context Protocol, allowing LLMs to programmatically manage layers, edit features, and execute spatial processing tasks.

A collection of Claude Code skills and playbooks that automate B2B cold email infrastructure, lead sourcing, and copywriting based on patterns from 1,000+ campaigns.

Foreman is a Python async runtime that supervises AI coding agents (Codex, OpenCode, Hermes, DeepAgents) using decision‑model checks (completion, verification, worker health, drift, human escalation). It observes worker events, asks a model (TypeSafe Jev or OpenAI) for probabilities, and deterministically directs the worker (continue, steer, stop, retry, verify, finish, or escalate). Install via `pip install foreman-core`; configure back‑ends and decision‑model adapters; run with `foreman run --repo <path> --job "<task>"`.

PocketPaw is a self-hosted, local-first AI agent platform that allows you to run and control your own intelligent assistants across multiple messaging apps and desktop interfaces without cloud subscriptions.

MCP PHP SDK is a framework‑agnostic library for building Model Context Protocol servers and clients in PHP. It provides attribute‑based tool/resource definitions, STDIO/HTTP transports, client helpers, and weekly conformance testing, enabling PHP applications to expose AI‑callable functions to LLMs.

A community-driven, open-source collection of study materials, practice problems, and resources for students preparing for the GATE Data Science and Artificial Intelligence exam.

A production-oriented AI workflow framework that organizes LLMs, scripts, and APIs into versioned, recoverable nodes to ensure stable and cost-effective execution.

ManimCat is an AI‑assisted web workspace that turns natural‑language math prompts into animated (Manim) or static (matplotlib) visuals. It offers a quick‑generation “Workflow” mode and a longer‑lived “Agent/Studio” mode for iterative creation, built with React, Express, Redis, and optional Supabase storage, and can be run locally or via Docker/Hugging Face Spaces.

A high-performance C++ port of NVIDIA's NeMo speech recognition models that enables efficient, Python-free transcription, speaker diarization, and sound tagging on CPU and GPU.

NVIDIA NeMo Agent Toolkit (NAT) is an open‑source Python SDK that adds enterprise‑grade instrumentation, profiling, and optimisation to LLM‑based AI agents. It works with popular frameworks (LangChain, CrewAI, Semantic Kernel, etc.) and provides a CLI (`nat`), a UI, protocol support (MCP, A2A), and experimental runtime intelligence (Dynamo). Install via `pip install nvidia-nat` and define agent workflows in YAML to run, evaluate, or serve them.

A guide repository for writing structured prompts for Seedance 2.0 (ByteDance's AI video model), with 16+ templates, 8+ examples, platform links, usage steps, workarounds, and hard limits.

Clawd‑Code is a Python re‑implementation of Anthropic’s Claude Code agent, providing a streaming REPL, tool‑calling loop, multi‑LLM provider support, and extensible markdown‑based “skills”. It ships with a full CLI, 30+ built‑in tools, tests, and documentation, and is installable via a simple virtual‑env setup.

zerostack is a lightweight coding agent written in Rust that runs in the terminal, supporting multiple LLM providers. It offers granular permission controls, session management, subagents, and optional persistent memory, while using far less RAM and CPU than JavaScript-based agents.

OpenPipe / ART is an open‑source Python library that adds reinforcement‑learning (GRPO) training to LLM agents. It offers a client‑server architecture (local or server‑less via W&B) that handles inference with vLLM, trajectory logging, reward assignment, and LoRA‑based policy updates. Ready‑made Colab notebooks demonstrate agents learning tasks such as email search, playing 2048, Tic‑Tac‑Toe, and Codenames. The library aims for low cost, fast scaling, and zero‑ops infrastructure, supporting most vLLM‑compatible models under an Apache‑2.0 license.

An unofficial, community-maintained "awesome list" that collects official documentation, tutorials, real-world field cases, plugins, reviews, open-source alternatives, and global meetup listings for Grok Bot — an always-on AI teammate that works on a shared cloud computer and keeps going after you close your laptop.

An operating system for AI agents that provides a persistent runtime environment with automation, supervision, and state management for ACP-compatible agent CLIs.

A family of multi-platform GUI agents and foundation models that automate tasks across mobile, desktop, and web interfaces using visual perception and planning.

CodeIsland is a macOS app that puts your AI coding agents in the MacBook notch, showing their live status, letting you approve tool calls or answer questions without leaving your current window. It works with 30+ coding tools and keeps everything local and private.

senpi is a TypeScript monorepo that rebrands pi-mono's coding agent as a saner, extension-first runtime with built-in intent gating, compaction, permission controls, and more. It serves as the runtime for the Dori AI assistant and offers a lightweight alternative to OMO as a single CLI binary.

docmd is an open‑source tool that turns a folder of Markdown files into a static documentation website **and** a suite of AI‑ready outputs (search index, LLM‑context files, Open Knowledge Format bundles, MCP server, and a RAG‑powered chat assistant). It works with zero configuration, supports plugins for SEO, search, OpenAPI rendering, etc., and offers a free Cloud Relay to proxy AI requests without exposing provider keys. Ideal for projects that need both human‑readable docs and machine‑readable knowledge bases.

TE MAN is a comprehensive ComfyUI plug‑in that adds unlimited‑canvas workflow, prompt‑enhancement nodes (Qwen, MiniMax H3), audio/video editing, a 3‑D director panel, an asset library and a chat‑assistant “构想台”. It works with local diffusion models and cloud APIs, targeting artists and prompt engineers who want an all‑in‑one UI for image, video and audio generation.

A desktop application that creates persistent AI characters with a continuous identity across text chat, a tabletop RPG engine, and an autonomous OS-level agent.

Awesome AI Pedia is a curated AI knowledge base and blog platform that centralizes practical guides on AI coding tools, MCP, agents, prompt engineering, and related topics. It offers a searchable, dark-themed website with auto-generated navigation and automatic deployment, helping developers quickly find and apply AI best practices.

An interactive app that demonstrates how LLMs work by walking through each stage of the pipeline, from a pattern-matching chatbot to training a GPT-style transformer from scratch with no ML libraries. Each section includes live code, visualizations, and references to the original papers.

chan.py is a Python 3.11 library that implements the Chan‑Theory technical‑analysis framework, offering calculations of strokes, segments, central zones, and buy‑sell points, plus feature engineering, ML model integration, back‑testing, and live‑trading connectors (Futu, data APIs, MySQL/SQLite). It includes plotting utilities for static and animated visualisation.

A framework for automated search and optimization of the code scaffolds (harnesses) that manage data and retrieval for fixed AI models, improving task performance without retraining.

The VLA Handbook is a continuously‑updated, open‑source knowledge base that collects, analyses, and distills Vision‑Language‑Action robotics papers. It provides deep‑dive theory markdowns, practical deployment guides, community‑sourced tips, industry radar, and an AI‑assistant skill, all driven by an automated Pulsar pipeline.

HackerAI is an agentic framework for building AI-powered penetration testing assistants that can perform security analysis in isolated cloud environments.

A non-official SDK and CLI for WeChat Reading that enables over 40 read-write operations, including book management, note synchronization, and public account feed exports.

free-code is a fork of Anthropic's Claude Code terminal AI coding assistant. It removes all telemetry (no data sent home), strips injected security-prompt guardrails (while keeping the model's own safety training), and unlocks 54 experimental feature flags. It supports five AI providers (Anthropic, OpenAI Codex, AWS Bedrock, Google Vertex AI, Anthropic Foundry), is built with Bun and TypeScript, and includes an IPFS mirror so the code persists even if the repo is taken down.

An inference engine and training environment for Recursive Language Models (RLMs) that enable LLMs to handle near-infinite contexts by programmatically decomposing tasks via recursive sub-calls in a REPL environment.

A ComfyUI node pack that adds instruction-based image editing to Krea 2 (Raw or Turbo) by injecting the source image both as VAE latent tokens and into the Qwen3-VL text encoder, matching how the accompanying LoRA was trained. It provides two nodes for model patching and grounded encoding, supporting single-image edits and two-input person-into-scene restaging with aspect-ratio fitting and VRAM-aware pre-encoding.

An autonomous business operating system for solo founders that provides AI agent skills to automate everything from idea validation and product building to marketing and 24/7 operations.

A lightweight inference engine for GPT-SoVITS that optimizes speech synthesis for high-speed CPU performance and low memory usage.

A collection of prompt engineering skills for Claude that generate professional, cinematic video prompts for Seedance 2.0 on Higgsfield.

A streaming-optimized Markdown renderer for React that handles incomplete AI-generated content gracefully to prevent formatting glitches.

Better Design is an MCP server that provides AI coding agents with professional design systems, component code, and accessibility review rules to generate high-quality, consistent UI.

Hermes Gadget is an open-source SDK and firmware for building a physical, push-to-talk voice interface for an AI agent like Hermes. It runs on ESP32-S3 and Raspberry Pi boards, includes a desktop simulator, and lets the agent control the device's LEDs and sensors as tools.

A local proxy that pools multiple AI subscription accounts and rotates between them to prevent interruptions from usage limits during high-volume coding agent tasks.

FastWAM is a high-performance World Action Model for robotics that enables fast action prediction by optionally skipping test-time future imagination, significantly reducing inference latency.

TalkingHead is a real-time 3D avatar engine for the browser. It renders a customizable GLB model, gives it speech and lip-sync (via TTS with timestamps), adds idle behaviors and mood changes, and integrates with external APIs like OpenAI's Whisper or audio-driven lip-sync modules. It's used in various AI-powered apps and research projects.

Interceptor is a local automation bridge that enables AI agents to interact with your existing browser sessions, native desktop applications, and physical iPhones to perform complex, real-world tasks.

An open-source framework for building real-time multimodal conversational AI agents with low-latency audio, text, and visual capabilities.

A large-scale open-source library for 2D/3D image and point cloud processing, used widely in robotics and 3D perception.

A Deep Reinforcement Learning library providing single-file, transparent implementations of popular algorithms to simplify research, debugging, and prototyping.

AxonHub is an AI gateway (Go-based, web dashboard included) that accepts requests in OpenAI, Anthropic, or Gemini SDK formats, then transparently translates and routes them to 10+ supported model providers. It adds request tracing, RBAC, load balancing, and per-request cost tracking — letting you swap models (e.g., GPT-4o to Claude) via config changes instead of code rewrites.

Osaurus is a native macOS app that lets you run private AI agents locally. Agents have their own memory, can execute code in an isolated Linux VM or macOS sandbox, call tools (file system, browser, Apple apps, image generation), and communicate securely via encrypted channels. The platform works with local models (MLX, Apple Foundation Models, Liquid AI) or cloud APIs (OpenAI, Anthropic, etc.), and includes privacy‑filtering, cryptographic identities, plugin support, voice input, and system‑wide shortcuts. Install via Homebrew or a .dmg; requires macOS 15.5+ on Apple Silicon.