Trending AI projects

An AI-assisted tool that generates storyboards and stitches them into coherent videos using Gemini and Vertex AI Veo.

ComfyUI LoRA Manager is a web‑based tool (http://localhost:8188/loras) that helps you browse, download, organise and apply LoRA models for Stable Diffusion within ComfyUI. It integrates directly with CivitAI, offers a Chrome/Firefox extension, lets you save LoRA “recipes”, supports wildcard prompts, and provides dynamic filename placeholders for the Save‑Image node. Install via ComfyUI’s Custom Node Manager, a portable package, or manually, then use the provided Lora Loader node to insert LoRAs into your workflows.

WorldGen is a tool that generates immersive 3D scenes from text prompts or images in seconds, supporting both Gaussian Splatting and mesh outputs for VR, games, and simulations.

tts‑bench is a Python‑based benchmark suite for local text‑to‑speech models. It lets you install dozens of open‑source TTS systems, run them on CPU, CUDA or Apple‑Silicon, and compare speed (TTFA, realtime factor, memory), listen to generated audio (default voice and zero‑shot cloning) and view objective scores (UTMOS, WER, SIM). A public voting arena supplies human‑preference rankings for “best sounding” and “best cloning”.

A curated list of engineering resources and GitHub projects focused on building robust, secure, and reliable infrastructure for autonomous AI agents.

Embedding Atlas is an interactive visualization tool for exploring large-scale embeddings and tabular data, offering automatic clustering, cross-filtering, and multimodal support for deep data analysis.

LMDeploy is a comprehensive toolkit for compressing, deploying, and serving large language and vision models, designed to maximize inference speed and hardware efficiency.

A Microsoft‑maintained catalog of AI‑agent plugins (winappcli, winui, windbg) that let Copilot, Claude Code, Codex, OpenClaw and OpenCode automate Windows app packaging, WinUI 3 scaffolding, and WinDbg debugging.

A specialized environment for running the 176B-parameter Qwen3.8-Flash-Next model on a single NVIDIA DGX Spark by offloading embedding tables to NVMe and patching vLLM for memory efficiency.

A family of real-time stereo matching architectures that achieve strong zero-shot generalization by using knowledge distillation, neural architecture search, and structured pruning.

A 100-line minimalist LLM framework that uses a graph-based abstraction to build agents, RAG, and workflows without the bloat of traditional frameworks.

Tool‑Eval‑Bench is a CLI‑driven benchmark that evaluates how well LLMs served via OpenAI‑compatible endpoints can select, parameterise and chain external tools. It runs 69 deterministic scenarios (plus optional Hard‑Mode cases), records pass/partial/fail scores, and also offers throughput, long‑context, speculative‑decoding and accuracy tests. Results are saved as Markdown reports and SQLite records, with a star‑rating system and safety‑violation caps.

VMark is a local‑first desktop editor for many plain‑text formats (Markdown, JSON, YAML, TOML, code, etc.) that renders each file in a format‑specific view and integrates directly with LLM assistants via the Model‑Context‑Protocol, letting AI read and edit the same files you work on.

An ultra-fast, multi-tenant graph database that uses sparse matrices and linear algebra to power Knowledge Graphs for Generative AI and agent memory.

A curated, actively maintained list of 64 AI-focused tools for building beautiful user interfaces, organized into five categories (skills for AI assistants, apps, MCP plugins, non-AI design tools, and learning resources), with editor's choice picks and a contribution invite.

A catalog of 28 structured reasoning frameworks for AI agents that provides clear, repeatable procedures for complex tasks like diagnosis, decision-making, and risk assessment.

A code-first video and motion graphics engine that allows AI agents to programmatically render high-quality 3D and 2D visuals directly on the GPU without headless browser overhead.

Agent Flow is a real‑time visual debugger for Claude Code and Codex AI agents. It streams execution events via a lightweight hook (Claude) or Codex rollout logs, and displays them as an interactive node‑graph with timeline, transcript, and heat‑map panels. Available as a standalone web app (`npx agent-flow-app`), a VS Code extension, or a dev server, it supports multi‑session tabs, JSONL replay, and runtime selection. Designed to make agent reasoning, tool‑call chains, and performance visible for debugging and prompt‑engineering, it ships with optional anonymous telemetry and is Apache‑2.0 licensed.

A community-curated index of decision models (AI that answers questions with probabilities instead of text), plus their APIs, tools, benchmarks, and research.

A curated collection of reusable "skills" for Jev‑based AI coding agents—modular prompts for triage, document lookup, evaluation, and action selection—plus 67 community demos, local model adapters, and 108 example scenarios to help developers plug these abilities into browsers, games, code‑review tools, and more.

The huggingface/course repo contains the source content for the free, open-source Hugging Face course, which teaches applying Transformers to NLP and other tasks using the Hugging Face ecosystem. The course is translated into many languages (most WIP), and the README explains how volunteers can contribute translations, how to generate Jupyter notebooks locally, and notes that new chapters are only written by Hugging Face authors.

An open-source, self-hosted gateway that connects phone lines and messaging apps to AI agents, allowing for private, AI-powered receptionists and customer communication.

SHARP is a fast monocular view synthesis tool that uses a neural network to generate a metric 3D Gaussian representation from a single image in under a second.

Open AI UGC is an open‑source, self‑hostable web app that lets users generate short vertical UGC video ads using multiple state‑of‑the‑art text‑to‑video models. It includes Google OAuth, a credit system with Stripe billing, async webhook handling for renders, and a dashboard of past creations. The stack is Next.js 16, React 19, Tailwind, Prisma, and Muapi.ai for the underlying video generation.

A tool that converts text into official Chinese government documents (DOCX) following the GB/T 9704-2012 national standard, designed as a skill for AI assistants.

A hybrid-attention model based on MiniMax H3 that significantly speeds up video generation while maintaining high visual quality.

FlyEnv is a cross‑platform desktop app that natively installs, versions, and orchestrates runtimes (PHP, Node, Python, Java, Go, etc.) and services (databases, caches, search, web servers). It offers per‑project workflow definitions, local HTTPS domains, and an MCP server that lets AI coding assistants query the local stack. Available via installers for Windows, macOS, and Linux, it’s open‑source (BSD‑3‑Clause) and targets developers who want a native alternative to XAMPP/Laragon/Docker for full‑stack local development.

A cinematic visual direction framework for LLMs that transforms scripts and ideas into precise, consistent prompts for AI image generators.

PentesterFlow is an open‑source terminal assistant that lets a human analyst guide an LLM‑driven agent through penetration‑testing tasks. It integrates real security tools (curl, shell, Burp, file utilities), supports many LLM backends, stores evidence and findings as markdown, and retains project‑level memory across sessions while requiring explicit approval for risky actions.

OntoBricks is a Databricks-native app that turns Unity Catalog tables into a living, queryable knowledge graph using LLM-driven ontology generation and mapping. It provides governance, multiple graph backends, and exposes the graph to AI agents via MCP.

ai-usagebar is a system monitoring tool that displays real‑time quota usage for a wide range of AI coding services (Claude, Codex, GitHub Copilot, OpenRouter, DeepSeek, and many more) as a desktop widget, terminal UI, or JSON feed. It reuses existing credentials from the official CLIs, automatically refreshes tokens, and provides threshold alerts and reset timers.

Mirobody is a self-hosted AI health data engine that standardizes medical records, lab reports, and wearable data into a unified format, allowing users to query their health history with traceable citations.

A collection of 164 specialized skills for Claude Code that guide developers through the entire process of building, launching, and growing apps for iOS, macOS, watchOS, and visionOS.

NodeTool is an open‑source, agent‑first creative workspace that lets you build multimodal AI‑driven media (text, images, video, audio, 3‑D, games) via a node‑graph interface. Connect any LLM or generative model with your own API keys or run local open‑weight models, let the built‑in agent wire and run workflows, and edit the results in specialized editors (storyboard, timeline, sketch, 3‑D, game). Available as a free cross‑platform desktop/web app, extensible with custom TypeScript/Python nodes, and usable from a CLI or MCP server.

ProxmoxMCP-Plus is a bridge that lets AI agents (via MCP) and standard HTTP tools (via OpenAPI) safely operate a Proxmox VE virtualization server, covering VM/LXC lifecycle, snapshots, backups, ISOs, container commands, and persistent job tracking. It emphasizes security with API tokens, command policies, approval gates, and TLS validation, and supports local stdio, Docker-based HTTP, and an opt-in sandboxed Code Mode. An optional OAuth browser-gate flow backed by PostgreSQL handles remote client authentication.

Numen is an open‑source Minecraft mod that embeds an embodied AI agent controllable by large language models. Users install the mod, configure an OpenAI/Anthropic API key, summon a bot, and interact with it via natural language, voice, or scripted Lua actions. The project provides a three‑layer architecture (Minecraft core, Lua action API, LLM sandbox) and a plugin system for extending the agent’s capabilities.

fff is a Rust‑based, in‑memory file‑search library with fuzzy, frecency‑aware indexing. It powers a Neovim plugin, an MCP server for AI agents, and a Pi extension, offering ultra‑fast path and content searches that are cheaper for LLMs than traditional greps.

claude‑real‑video (crv) is a Python CLI / web tool that extracts smart keyframes and Whisper transcripts from any video (YouTube, local file, etc.) on your own machine. It deduplicates near‑identical shots, optionally adds speaker labels, and outputs frames, timestamps, and a transcript that you can paste into Claude, ChatGPT, Gemini, or other LLMs. The package also provides a local knowledge‑base (`crv‑ask`) and integrates with Claude Code, Cursor, Codex, and other agent platforms via a one‑line skill install.

TopDeepLearning is a curated, star-ranked list of popular GitHub projects related to deep learning. It's a single table (Project Name, Stars, Description) sorted by descending star count, last updated 2026.07.14. It covers frameworks, LLM tools, AI agents, vision, speech, and educational resources — with no code, instructions, or analysis beyond the listing itself.

IWE is a shared memory system for AI agents and humans that uses a folder of Markdown documents to provide persistent, editable, and searchable project context.

An extension of veRL for training LLM agents via reinforcement learning, featuring a step-independent rollout mechanism to efficiently handle long-horizon, multi-turn tasks.

An OBS Studio plugin that uses neural networks to remove portrait backgrounds and enhance low-light video in real-time without requiring a physical green screen.

gemini-web-to-api is a Go-based proxy server that exposes Google Gemini's web interface as standardized REST APIs. It lets you use Gemini's latest models without an API key by authenticating with browser cookies, while providing OpenAI, Claude, and Gemini-compatible endpoints.

An open-source agent harness engineering platform that uses OpenTelemetry to trace, evaluate, and guard AI agents and coding agents.

A JavaScript/TypeScript framework for building multi-agent workflows, supporting text, sandbox-based filesystem agents, and low-latency realtime voice agents.

An open-source framework for building full-stack AI applications with unified APIs for multiple model providers and cross-language SDKs.

A framework for automating the fine-tuning and pretraining of large-scale AI models with PyTorch-native parallelism and optimized recipes.

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.