Chat with an assistant that can read your hosts, snippets, fleets, alerts, automations, workspaces, network graph and homepage.
The assistant suggests changes, and nothing is applied until you approve it.
Can run read-only diagnostic commands on a host if you allow it.
Open the assistant from the terminal toolbar with the current session as context.
Works with Ollama, Anthropic, OpenAI, Google Gemini and any OpenAI compatible provider.

Settings

AI assistant: let users turn the assistant on. While this is off it is hidden for everyone.
Allowed private AI hosts: hosts on your private network a provider may point at, such as a self-hosted Ollama.

Enable the AI assistant: turn the assistant on for yourself.
Allow read-only diagnostic commands: let the assistant run safe commands on your hosts.
Providers: the AI providers, models and API keys you use.

Permissions

ai.use: Open the assistant and ask it things. Admins and users have it by default.
ai.manage_providers: Add, edit and remove AI providers and their API keys. Admins and users have it by default.
ai.apply_proposals: Let the assistant carry out the changes it suggests. Admins and users have it by default.
ai.services.use: Let other features call the assistant for you. Admins and users have it by default.
ai.secrets.share: Let other features use the stored AI API key without seeing it. Admins and users have it by default.

Uses from other plugins:

workspaces.saved, network-topology.graph, snippets.access, fleets.access, terminal.history, automations.access and homepage.items to read your data. Each one is optional.

npm run build      # build into dist/
npm run test       # run this plugin's tests
npm run typecheck  # type-check this plugin