Camera·

grab a front/rear frame, ask the vision model a question about it, screenshot the dashboard — tools/rover_camera.py, 2 tools.

From the module docstring

📷 Earth Rover camera tools — inline Converse image returns.

tool	does
`rover_see`	Look through the rover's cameras.
`rover_screenshot`	Capture composite rover views (front/rear/map) via SDK /screenshot.

`rover_see`·

rover_see(camera: str = 'front', save: bool = False) -> Dict[str, Any]

Look through the rover's cameras. Returns frame(s) inline so you can SEE.

Uses the SDK /v2 frame endpoints (fast, no disk I/O on the SDK side).

argument	meaning
`camera`	"front" (default), "rear", or "both". Rear camera only exists on "zero" bot types — if unavailable, the tool degrades gracefully to front only.
`save`	Also write JPEG(s) to ./captures/ and report the path(s).

Returns

Dict with status and content: text summary + inline image block(s).

rover_screenshot(views: str = 'front,rear,map') -> Dict[str, Any]

Capture composite rover views (front/rear/map) via SDK /screenshot.

Slower than rover_see (renders via headless browser screenshots) but includes the MAP view with the rover's position.

argument	meaning
`views`	Comma-separated subset of: front, rear, map.

Returns

Dict with status and content: text summary + inline image blocks.