Configure the environment
Project-local voice credentials, task-model selection and optional desktop connections.
The project .env
Voice commands require a project-local .env. Its GEMINI_API_KEY and JARVIS_* values replace inherited environment values with those names. Restart the voice session after changing its configuration; an active session retains the settings it started with.
GEMINI_API_KEY=replace-with-your-keyThe high-level task model
Use jarvis model to configure the provider and model used for delegated tasks. Supported custom endpoints and local servers still need valid configuration. Provider credentials, context limits and image/tool support vary.
uv run jarvis modelMicrophone and speaker selection
List available device indices before overriding the input or output device. Device names and indices depend on the current machine. The low-level client accepts input-device and output-device options for targeted diagnostics.
uv run jarvis lowlevel --list-devicesAdd only the connections you need
The browser setup command prepares an unpacked extension and native host; load the returned extension directory manually in Chrome or Edge. The editor setup installs the VS Code adapter. Native notifications also require Windows to allow popups.
uv run jarvis desktop setup-browser
uv run jarvis desktop setup-editor --install
uv run jarvis desktop setup-notificationsWhen a voice function is missing
Inspect the enabled functions printed at startup. A project JARVIS_LOWLEVEL_FUNCTIONS override can hide shipped function groups. Remove an obsolete override to use the current defaults, then restart the voice session. Use desktop tools to inspect the current schemas instead of guessing operation names.
uv run jarvis desktop tools