-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Draggy is a desktop AI assistant that talks, browses and works with your files, all on models running on your own computer. It drives a local Ollama instance, and nothing leaves the machine unless you ask for something that needs the internet: a web search, a page to read, a model to download.
This wiki covers using it and working on it. If you just want to install it and start talking to something, Installation is the page you want. The site is at draggy.org.
- Installation: requirements, getting Ollama, the first launch
- Choosing a Model: what fits your hardware, and how to change it
- The Chat Window: the composer, slash commands, attachments, history and export
- Tools and Capabilities: web search, files, running code
- Browser and Ad Blocking: the browser window, and what it blocks
- Document Library: searching your own files
- Voice Mode: talking out loud, and the settings that matter
- Extensions: borrowing tools from MCP servers
- Settings: every panel, in order
- Privacy and Data: what reaches the network, where files are kept, how to back up
- Troubleshooting: when something is wrong
- Building from Source: development setup and the test suite
- Architecture: how the pieces fit together
- RELEASING.md: how a version reaches people
- CONTRIBUTING.md: house style, and what a pull request means for copyright
Draggy is free software under the GNU GPL v3.0 or later.
It is not a wrapper around a cloud API. There is no account and no key, and there is nowhere for your conversations to go. The one exception is an extension you switch on yourself, which talks to whatever service it is for. Nothing is switched on by default.
It is also not magic. A model small enough to run on a laptop is a model that knows less and reasons worse than one running in a datacentre. Draggy tries to be honest about that rather than hide it: it will tell you when it is guessing, and it will search the web rather than invent a version number.