leon-ai/leon

Leon is an open-source personal assistant that you self-host on your own server. Think of it like a private version of Siri or Alexa, one you control completely, with no data sent to third-party services unless you choose to. You can talk to Leon or text him, and he responds by voice or text. Because it runs on your own machine, it can also run fully offline, which is the project's key privacy promise. The core idea is a "skills" system: each skill is a small bundle of code that teaches Leon how to handle a specific type of request. This means anyone can build a new skill and share it, so the assistant grows through community contributions rather than being locked to whatever the original author built in. The README describes this as "one core to rule them all", you add capabilities as skills rather than building separate tools for each idea. Under the hood, Leon figures out what you mean using a natural language understanding system (NLU), software that interprets sentences into recognized intentions. The project is working toward eventually supporting large language models (LLMs, the same technology behind tools like ChatGPT) for cases where the intent system can't match what you asked. The server and web app are written in TypeScript. Installation is done via a command-line tool called the Leon CLI, and a Docker option is also available. It requires Node.js version 16 or higher. The project is MIT-licensed and openly welcomes contributions.