Repository index
6791 AI, ML and Web3 projects · page 64 of 68
- Local-First Android AI Agent
An autonomous on-device AI agent for Android that automates OS tasks via accessibility services, continuous voice interaction, and self-evolving workflows.
- Curated Self-Hosted AI Tools
Weekly health-checked list of open-source AI for local LLMs, chat UIs, agents, RAG, and media generation. Dead projects auto-removed.
- Hands-On LLM Full-Stack Course
Comprehensive tutorial implementing LLMs from scratch: builds Transformer in PyTorch, reproduces Pretrain/SFT/LoRA/DPO/GRPO on single GPU, covers PEFT/Agent/RAG with real experiments and scripts.
- 🤖 ANDREJ-KARPATHY-SKILLS
Skill file for Claude Code based on Andrej Karpathy's experience with LLM assistants.
- ChatGPT Control SDK
An unofficial TypeScript toolkit that lets Codex agents control visible sessions of the ChatGPT web interface. It simplifies dialogue automation and integration with external systems.
- LOOMFEED - AGENT ALTERNATIVE TO REDDIT
An open platform for dialogues between AI agents and humans with provenance tracking, reputation, and epistemic status labels.
- AI for Test Automation
AI framework that explores test paths using AI and generates reproducible scripts for 30 times faster execution.
- CODEX LONG TASK MANAGEMENT
Reliable MCP control panel for long Codex Desktop tasks. 202 stars on GitHub.
- MCP OBSERVATORY - PROTECTION ON CREST
CI-native security testing for MCP servers: attack simulation, schema deviation detection, and health assessment before agent launch.
- Field-Level Data Pipeline Assurance
Mapping-driven source-to-target validation with fault-injection scoring and risk-adaptive test prioritization for data pipelines.
- Automated Twitch Drops Farmer
Background miner that auto-claims Twitch drops, switches channels, and collects timed rewards without downloading video. Runs on Windows, macOS, and Linux.
- FinSight-AI — Finance with AI
AI agent for stock analysis: fault-tolerant pipelines, RAG via pgvector, report versioning, and full evidence traceability.
- Engram: Agent Memory
Hybrid long-term memory plugin for OpenClaw agents, combining SQLite with FTS5 for structured facts and LanceDB for semantic search.
- File Routing for Workers
File-based routing framework for Cloudflare Workers. Routes are created automatically.
- PentesterFlow Agent
AI-driven agent for autonomous pentesting directly in the terminal. It enables Offensive Security tasks via natural language.
- VoxCPM2 TTS for ComfyUI
The VoxCPM2 module adds speech generation in ComfyUI for 30 languages with voice settings, voice cloning, and LoRA adapter training for personalization.
- One-Click Video Capture
The extension adds a button to the toolbar, allowing you to download videos from the current page to ~/Downloads with one click via a local yt-dlp daemon, supporting more than 1800 sites.
- CLICD — Container Control Panel
CLICD provides a lightweight web panel for managing LXC and KVM virtualization: web console, batch operations, image management, NAT networks, IPv6 allocation, WebSSH, VNC, resource control, and security.
- Pi Knock Notifier
Send push notifications via Pushover, ntfy, or webhook when your Pi coding agent finishes a task or requires input.
- nl-veil: AI CODER IN ZIG
nl-veil — open-source AI coding assistant with parallel agents and persistent project memory. Works with local models, Ollama or cloud LLMs on Windows, macOS and Linux. Written in Zig, it already has 201 stars.
- Affinity Designer 2 Desktop Build
C++ desktop application bundling Affinity Designer 2 with vector editing, raster export, color management, and portable mode for Windows 11.
- Local Claude Design Agent
Run Claude Design locally as a skill agent in Cursor, Claude Code and other environments, creating ready-made UI mockups, prototypes, slides, and wireframes in pure HTML without calling claude.ai/design. Best with Opus 4.8.
- SakikoMind AI Agent
AI agent platform for customer service with hybrid intent recognition, multi-agent orchestration, MCP-RAG, dynamic skills, memory, and human handoff.
- Lightweight LLM Chat UI
A minimal LLM chat interface with two input fields, under 100 KB and zero dependencies.
- AI Content Generator
Python tool with AI models for generating realistic uncensored images and videos from text prompts.
- Claude Code Backdoor
The project lets you run Claude Code on any AI provider: DeepSeek, Groq, Ollama, and OpenRouter through a single Python interface.
- LLM TCO Calculator
Bindwidth estimates hardware requirements and TCO for running LLMs locally, using real data to calculate required GPU/CPU and energy consumption, depreciation, and licensing costs.
- Agentic Inference Framework
openJiuwen-ai/agent-infer is a vLLM-based framework for efficient execution of agentic workloads. It simplifies deployment and scaling of LLM inference in multi-agent scenarios.
- Open Turkish TTS Suite
Ahenk provides an open Turkish text‑to‑speech system featuring the VoxCPM2‑LoRA model, an 8.6 M‑parameter on‑device version, and a Turkish text normaliser.
- SheetasToken — Table Data Analysis
Graph-based Python framework for searching and analyzing data in multi-page tables. Simplifies working with linked data.
- AI Trip Planner
The project implements a model-agnostic trip planner built on LangChain, LangGraph, FastAPI, and Vue. It can generate personalized routes based on user preferences and external data.
- AI-SpicyLabs Generator
A simple tool for creating photorealistic images and videos using AI. No complex installation — just get high-quality results instantly.
- Among Us Premium Assistant
AI assistant and modding framework for Among Us on Windows 2026, providing real hints, custom content, and session analysis.
- iOS Jailbreak Tweak Best Practices
Agent skill providing best practices and guidelines for developing iOS jailbreak tweaks using the Logos framework.
- Bhagavad Gita Knowledge Graph
Explore all 700 verses in 3D with 20 commentaries, audio recitations, GraphRAG chat, and private journal — fully offline.
- Cluster Performance Monitoring
Kubernetes scheduler for HPC, ML and CI/CD with NUMA optimization on EPYC/Threadripper. Reduces latency on 128+ core nodes.
- SwiftUI shaders for UI
Library of 41 ready-made Metal effects for SwiftUI: holographic foil, kaleidoscope, magnetic fields, glitch, neon edges, etc. Integrates via typed View modifiers with full documentation.
- AI Starter Kit
A collection of ready-made AI agent skills, checklists, templates, and guides for Swiss teams, excerpted from the author's blog.
- ORCHESTRATING CLAUDE AGENTS
Mavericks-os orchestrates Claude Code sub-agents in long-running tasks. Stores state in artifacts, not chat history. Supports multi-repo, dependencies, and QA gates.
- 21-Day AI Dialogue
An AI existential-psychology consultant asks three questions each evening in Russian and English, remembers the answers and returns them as a reflection, helping the user understand themselves better in 21 days.
- Library Management System
Educational library system project with four roles: administrator, librarian, student, and visitor. Implemented in pure PHP using mysqli and prepared statements, without frameworks or build steps.
- Ridge Racer V macOS Port
Fukami statically recompiles Ridge Racer V (USA) to run natively on macOS and Steam Deck, requiring only the game's executable and no original data.
- Campus Knowledge Agent
FastAPI platform for campus document parsing, time‑range Qwen3‑RAG retrieval, evidence tracing, and manual review.
- Local AI Film Studio
Run a full AI film studio locally with recurring characters, video generation, voices, music, and a timeline editor. Includes 214 MCP tools for AI agent control.
- OpenRath: Multi-Agent Framework
Open-source framework with a PyTorch-like API for building dynamic multi-agent workflows in Python.
- NOSIS — LIGHTWEIGHT AGENT RUNTIME
Nosis is a personal agent with OS-level sandboxing, computer use, and built-in tools. Sessions, plans, and memory persist to files. Single runtime for GUI and TUI.
- WorkMesh: AI Employment Platform
Open-source platform that unifies skills, vacancies, enterprises, governance, and AI agents. It automates recruitment and ensures regulatory compliance.
- AI Image and Video Generator
Try AI tools for creating realistic images and videos without limits. Generate content quickly and without censorship.
- ReadyLLM Visual Platform
Deploy, configure, and monitor LLMs via an intuitive web interface, and also generate images and video thanks to built-in integration with ComfyUI.
- YuE2 ComfyUI Music Toolkit
Provides ComfyUI nodes for the YuE2-3B model enabling piano‑roll editing, track retakes, audio/MIDI cover generation, lyric writing, and LoRA integration.
- Self-Hosted AI Coding Remote
Control Claude Code and Codex agents across all your machines from a browser or phone. One Docker container, full terminal/file/git access.
- Local Translation of Foreign Videos
The project lets you automatically add Chinese subtitles to a foreign video with one AI prompt: download, recognition, translation, polishing, and embedding — all locally, without paid transcription APIs.
- Agent Fleet Manager
Engine for scalable, reproducible data collection using a fleet of autonomous worker agents. It allows easy configuration of data sources, fault tolerance, and result export.
- Burrow — GUI for Mole
Free open-source interface for Mole CLI on macOS: cleanup, deletion, optimization, disk analysis and real-time monitoring, plus long-term history and MCP server for Claude Code.
- CapCut Pro Studio 2026
CapCut Pro Studio 2026 offers professional video editing with 4K HDR export, AI transitions, and a full resource library without limits.
- Agent-Py AIOps Toolkit
A personal AIOps framework that implements durable workflows, scoped hybrid retrieval, reliable event delivery, and verified API/browser contracts in Python.
- Time‑Aware Chat Narrative Framework
Provides a persistent, time‑aware narrative layer for Koishi chatbots, enabling branching storylines that remember past interactions.
- LLM Token Savings via CLI
A CLI tool that delegates routine tasks to cheap LLMs (Kimi, DeepSeek, Ollama), saving 60-70% of the token budget.
- SX1262 Chirp Signal Generator
The project demonstrates using the LoRa SX1262 chip to create coherent linearly frequency-modulated chirp signals through a simple HTML interface. Users can set parameters and observe the signal waveform in real time.
- DeepFake AI Kit
LLM-based toolkit for detection, analysis, and responsible creation of synthetic media in real time.
- Agent Campus Simulation
The project implements a multiplayer social campus simulation using LangGraph, providing memory, document search (RAG), personality evolution, and reliable task execution.
- INTENT-ROUTER INTENT COMPILER
Intent-Router turns vague agent requests into typed IntentSpec contracts (probe, ask, halt) before routing. It serves as the input layer for Jev and Laya models, guaranteeing precise understanding of intentions.
- Task-Oriented AI Workbench
An AI workbench where agents autonomously select tools like search, web reading, maps, and knowledge bases in a single loop, with fully traceable sources.
- Dark Souls PS2 Homebrew Port
A homebrew PlayStation 2 port of Dark Souls Remastered built with a custom C/VU1 engine. Includes asset exporters that extract data from your own game copy.
- AI Image and Video Generator
A C# toolset for creating and generating images and video using AI models, aimed at creative projects and content.
- GB10 Read-Only OS Builder
Shell scripts that build a minimal read-only OS image for GB10-based DGX Spark and ASUS GX10 boxes.
- AI Inference Gateway
Synapass is a Go‑based AI inference gateway that routes requests across multiple LLM providers, applies policies, caches responses, and provides observability.
- Local Stable Diffusion Reference
Curated notes on samplers, LoRAs, and VRAM optimization for running Stable Diffusion locally.
- CapCut Pro Studio
High-performance video editor with support for 4K HDR export, AI transitions, and a full library of professional resources without limits.
- Ubuntu Voice Image for ThinkSmart View
Pre-built Ubuntu image for Lenovo ThinkSmart View (CD-18781Y) that boots directly into a voice assistant UI, with flashing guides and build scripts.
- Stable Diffusion Local Guide
Curated notes on samplers, LoRA training, and VRAM optimization for running diffusion models locally.
- Local Memory for LLM
Mnemo provides a local memory layer that stores knowledge as an entity graph, extracts facts, and enables semantic search. Compatible with Ollama, OpenAI, Anthropic, and any OpenAI-compatible API.
- LLM Intern Skill
This project provides a set of LLM-based tools for improving resumes, matching jobs, fact-checking experience, and preparing for interviews.
- Trading Agents for A-Shares
Multi-agent framework with 7 analysts and risk assessment for the Chinese market.
- WorkMesh AI Employment Platform
WorkMesh is an open AI-era platform that connects skills, jobs, enterprises, management, and AI agents for efficient employment.
- DeepSeek V4.1 on DGX Sparks
Runs DeepSeek-V4.1-Flash across two NVIDIA DGX Spark systems, featuring EXL3 routed experts, 262k token context, 2M‑token KV cache, vision, tool use, MTP, and Pi bundling.
- Telegram AI Expense Logger
Telegram bot that converts natural-language messages into structured Google Sheets entries via LLM parsing.
- Free Roblox Studio AI Agent
Free Windows AI agent that plans, writes, and playtests Luau code in Roblox Studio with live diffs.
- AI Search in CLI
Open-source CLI tool for integrating AI search, recommendations, and conversational search into agent and business systems.
- Local AI Tool Call Verifier
Tiny 197 KB Python tool that semantically validates AI agent tool calls locally. Exit codes alone don't prove correctness.
- Love2D Game Engine Framework
A 2D game engine built on Love2D with three abstraction layers, providing tools without enforcing a specific architecture.
- Love2D Live Reloader
Automatically reloads Love2D games on file changes with built-in Moonscript compilation support.
- Ghostty Terminal Cursor Animations
Thirteen GLSL-powered animated cursor effects for Ghostty terminal, including special Enter and done animations for Claude Code and zsh integration.
- Love2D Grid UI Framework
LUIS is a flexible GUI framework for Löve2D enabling grid-based, layered interfaces for games and apps.
- AI-Powered Database Telegram Bot
A Telegram bot that lets you query MySQL, PostgreSQL, and SQLite databases using natural language via local LLMs with tool calling, SQL auditing, and Docker support.
- Interactive AI Telegram Bot
A Python Telegram bot using DeepSeek API, RAG web search, and PostgreSQL, with built-in smoke tests.
- InfiPlot AI Drama
InfiPlot is an interactive game where AI creates the plot and illustrations in real time. Player choices instantly shape the story.
- Time Series Architecture Atlas
The project collects modern time series forecasting models with concise documentation and working code to start research quickly.
- Ruach Studio Music Workbench
Ruach Studio provides a WebUI‑based music suite for YuE2, enabling score creation, LoRA training, stem separation, remastering, upscaling, lyric verification, and DAW export—all on your GPU.
- Cross-sector AML compliance framework
EvoAML framework combines graph networks and temporal evolution analysis for cross-sector tracking and BSA/AMLA 2020 compliance.
- AI Agent Platform
A platform for creating role‑specialized AI agents with support for multiple LLMs, MCP, and task management via a web interface and CLI.
- Elastic Data Loading
Zephon offers fast, flexible, elastically deterministic data loading for foundation‑model training, ensuring reproducible sampling and high throughput.
- Offline Tourist Safety
The heard project implements an offline group safety network for hikers based on ESP32, GPS, and LoRa, including a firmware‑in‑the‑loop simulator and 3D trajectory viewer.
- AI Web Answer Crawler
A MCP server and CLI that crawls websites to locate answer pages for AI agents, achieving twice the hit rate of Crawl4AI at one‑tenth the cost of Claude crawling.
- Pi Documentation Navigator
The project provides tools for navigating documentation and creating an agent-readable playbook for building based on earendil-works/pi.
- Skill Extraction from Media
The project converts videos, PDFs, transcripts, and notes into ready-made skills for AI agents, preserving source links.
- NaturePanelForge: code-first
NaturePanelForge turns scientific figure images and open Nature-family articles into tasks for recovering executable chart-building code at the individual panel level.
- Browser city builder on Three.js
rawprogress/fable-cities — a browser-based 3D city simulator, a Cities: Skylines analogue, built with Three.js and fully generated by AI agents. The source code will be released after development is complete.
- Dual RAG Bioinformatics Tools
Provides a dual retrieval‑augmented generation framework that recommends suitable bioinformatics tools for user queries.
- Offline Hairstyle Gallery for ComfyUI
Browse 893 mannequin hair references offline in ComfyUI with bilingual UI and Qwen Image 2.1 workflows.