TypeScript multi-agent framework — one runTeam() call from goal to result. Auto task decomposition, parallel execution. 3 dependencies, deploys anywhere Node.js runs.

A powerful Zotero AI and MCP plugin with ChatGPT, Gemini 3.1, Claude, Grok, DeepSeek, OpenRouter, Kimi 2.5, GLM 5, SiliconFlow, GPT-oss, Gemma 4, Qwen 3.5

Gemma Gem runs Google's Gemma 4 model entirely on-device via WebGPU — no API keys, no cloud, no data leaving your machine.

🚀 Pytorch Distributed native training library for LLMs/VLMs with OOTB Hugging Face support

Local AI Assistant on your phone

PokeClaw (PocketClaw) — first on-device AI that controls your Android phone. Gemma 4, no cloud, no API key. Poke is short for Pocket.

Local AI desktop app — chat, agent mode, image gen, video gen. Supports Ollama, Gemma 4, Llama, Qwen, OpenAI, Anthropic. Single .exe, no Docker.

Interactive CLI for creating & running multiple agents across CLI, Slack & WhatsApp; 100% local, thanks to Ollama & Gemma 4

Fast Gemma 4 inference in pure Java

A C# inference engine for running large language models (LLMs) locally using GGUF model files. TensorSharp provides a console application, a web-based chatbot interface, and Ollama/OpenAI-compatible HTTP APIs for programmatic access.

SAM3-Gemma4-CUDA is an experimental computer vision and multimodal reasoning application that combines Facebook's Segment Anything Model 3 (SAM3) with Gemma 4 multimodal model. This suite offers a comprehensive interactive web interface for high-precision image object detection, video segmentation, and interactive click-based tracking.

Experiments running offline LLMs in Python and Rust locally using Ollama and llama.cpp

The YAML standard for AI agent governance. Write rules once, compile to any runtime. Deterministic. Sub-microsecond. Three templates: enterprise, DAO, AI safety. Seven governance laws.

Awesome Gemma 4

100-question benchmark comparing proprietary (2025) and open-source (2026) LLMs.

LiteRT-LM for .NET

🔥 Python / Mojo Interface for Google Gemma 4

Local, AMD-supported autonomous browser agent. Privacy-first task automation via LM Studio and Gemma 4.

Dockerized inference server and benchmarks for Gemma 4 26B on the NVIDIA DGX Spark (GB10). Features ARM64 CUDA 13 builds using llama.cpp.

Go TurboQuant using GoMLX (Work in Progress)