Run a private local AI model on your own Mac or Windows computer without cloud API costs.
Let an AI agent read, edit, and run commands in a project folder inside a sandboxed environment.
Connect a custom OpenAI-compatible endpoint instead of using Zeraix's built-in local models.
| zeraix/zeraix | jwm0203/meetingcopilot | simular-ai/simulang | |
|---|---|---|---|
| Stars | 129 | 129 | 128 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 3/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs 16GB RAM recommended (8GB minimum for small models) and several gigabytes of disk space for model downloads.
Zeraix is a desktop application that lets a person run AI language models locally on their own computer, rather than relying on a company's cloud servers. It is described as a workspace where a user can chat with models, manage files, use built in tools, and run AI agents, all without needing an account, subscription, or API key for the core local features. The project has two parts. The first is Zeraix Desktop, the open source application anyone can download and use today. It manages GGUF format models, recommends models and settings based on a person's hardware, and can run agent-style commands inside an isolated QEMU sandbox for safety. It can also connect to a person's own OpenAI-compatible API endpoints or optional hosted models. The second part, called Zeraix Model Systems, is ongoing research the team is doing into running AI models more efficiently on regular consumer hardware, covering things like memory use, decoding speed, and support for different model architectures. The runtime technology coming out of that research, called ExactFlux, is not yet included in the public code because it is still changing quickly, the public app currently relies on the existing open source llama.cpp runtime instead. The team says ExactFlux will be open sourced once it reaches a stable point, and for now they will share research progress without presenting it as a finished feature. To get started, a person downloads a macOS or Windows installer, installs it, then opens Zeraix's Model Library, which checks the computer's memory and graphics hardware and suggests a model to download and run. Once a model shows a Running status, the user can start chatting with it directly on their machine. An optional Developer Mode lets Zeraix read and edit files in a chosen project folder, run terminal commands, and use browser tools, with the README recommending that a user keep projects under version control and review file changes and commands before approving them. Zeraix is released under the AGPL-3.0 license, a copyleft license that requires modified versions, including ones run as a network service, to also be shared under the same license.
A desktop app for running AI language models locally on your own computer, with an optional sandboxed developer mode for agent tasks.
Mainly TypeScript. The stack also includes TypeScript, llama.cpp, QEMU.
Modified versions, including ones offered as a network service, must also be released under the same AGPL-3.0 license.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.