Drive and watch a running dimOS robot: a 3D view placed by TF, cameras, a 2D map, keys / gamepad / phone sticks, arm planning and an mcap recorder. A dimOS Desktop app, meant to be forked per robot.
dimos-desktop install https://github.com/dimensionalOS/dim-controllerOpen Controller from Desktop's rail while a blueprint (sim, replay or robot) runs. Topics show up as they flow; nothing is configured by topic name. Details for every part: docs/reference.md. Changing the UI (adding a panel, a layer, a setting): docs/changing_the_ui.md.
Every topic is drawn by its message type (point clouds, odometry, paths, markers, detections, costmaps, TF, cameras in
3D) and placed by TF. A robot with a URDF on /robot_description is drawn from it, meshes and Gaussian splats included; a robot that
serves a USD stage instead (dimos's UsdManager) is drawn from that, the same way.
The layers button has every layer: each point cloud with its color, size and bandwidth, and the robot model's rows (URDF, Meshes, Textures, Splats, a row per URDF group).
Each camera is a panel; its gear trades latency for quality (Low latency, Balanced, High quality), done by the gateway.
The link chip opens Network: throughput and latency history, and what's using the link, stream by stream, with each stream's drops and its priority.
The rails become drawers, the view switch (top left) picks the main view, and driving is two thumb sticks with STOP between them. On its side the main view fills the screen.
WASD / arrows drive while held (Q/E strafe, Shift boosts), Space always stops. A gamepad works too (LT + RT stops).
There's no arming: every command goes through the gateway with a deadman, and driving disengages when latency passes
Settings → Drive → Max latency. The controller format (dog, humanoid, wheeled, arm, drone, or auto) sets the keys and
speeds. / opens the palette (every action), ? lists every key.
With a manipulation blueprint running, the arm view does what viser does: drag a gizmo, move joint sliders or pick a saved pose, and the planner checks it (orange: reachable, red: not); Preview plays the plan as a ghost, Execute runs it.
Record writes an mcap of every stream (dimos's logs included), readable in Foxglove. … picks the folder, leaves out the biggest streams, and sets image format and per-stream rates.
Every action is an HTTP endpoint the page itself uses, listed in the served /openapi.json (server/src/api.rs,
mirrored in dimos.yaml's provides:; deno task check-endpoints [--write]). Settings live in the backend, so a
change from the agent or another window shows in every open page.
Everything robot-specific is one file, a profile under frontend/src/profile/ (dog.ts, humanoid.ts, wheeled.ts,
arm.ts, drone.ts): the robot's frame, topics, speeds, keys and extra controls. E.g. a torso height slider on R / F:
keys: {
...groundKeys,
KeyR: { control: "torso_height", step: 0.02 },
KeyF: { control: "torso_height", step: -0.02 },
},
controls: [{
kind: "slider", id: "torso_height", label: "Torso height",
topic: "/torso_height", type: "std_msgs.Float32",
min: 0, max: 0.4, step: 0.01, initial: 0.2, unit: "m",
message: (value) => ({ data: value }),
}],A new message type is one file in frontend/src/layers/ (copy the closest, change types, import it in
layers/index.ts).
cd frontend && deno task dev # vite on :5173, proxied to a Desktop on :7077 (DESKTOP_URL)
deno task typecheck && deno task test # types; the tests (deno)
cd ../server && cargo test # every endpoint; the recorder and driving over zenoh on loopback
deno task check-endpoints [--write] # dimos.yaml's provides: = the served openapi.json
nix build .#dimosApp # what Desktop builds: bin/dimos-app-server serving the pageScreenshots (docs/screenshots.py): dim-go2-sim's ranch scene with two people, and xarm-perception-sim.
Robot-type icons (frontend/public/robots/) are modified from: "Humanoid robot" and "robot dog" by Izwar Muis and
"Robotics" by rukanicon (Noun Project, CC BY 3.0), and Tabler Icons
"drone" (MIT). Links in docs/reference.md.
Scene: "Home Scan" by luxury_scans (https://superspl.at/scene/3f89bbd3)















