AI features
The MythBrowse AI home, the Myth panel, your own model, and the Myth agent.
MythBrowse's assistant, Myth, talks to your own model: a local Ollama server or any OpenAI-compatible API with your key. AI output can be wrong.
Ask from the address bar
The address bar decides what you meant:
- a URL opens directly;
- any other text is sent to the AI home as a prompt;
- text starting with
?searches the web.
AI chat home
The new tab page is an AI chat, with saved chat history.
The Myth panel
Ctrl+J opens Myth beside the page. It can:
- summarize the page or pull out its key points;
- explain, translate or rewrite a selection;
- compare tabs, find things on the page and answer questions about it.
Sites you set to Don't share with Myth are never sent to your model.
Ask Myth to fix
When a page looks broken, Ask Myth to fix suggests a CSS / JS patch. You preview it, then keep or discard it.
Set up your model
Open the new tab page, then the model settings, and choose:
- Ollama: local, so nothing leaves your computer;
- OpenAI-compatible: a base URL, a model and your API key.
API keys are kept in the operating system's secure storage. See Privacy & data for what each request contains.
The Myth agent
In development builds only
The Myth agent (opening, reading and acting on web pages for you, with your approval) is proprietary and is not part of the open build. In the open build, Myth answers from your model only, and asking it to do something on a page gets a short notice instead.
Private development builds (v0.4.3, not released) also add AI translation, writing help and history you can search by meaning.