Interactive Environments and Physics
Telemus AI supports objects, models, mind maps and physics activities alongside ordinary workspace panels. Available controls depend on the object and host.
Selecting and moving objects
Use a controller, pointer or supported touch gesture to select an object. A visible highlight identifies the current target. Supported mouse layouts offer rotation and uniform-scale handles; touch can provide direct dragging and scale controls without every mouse handle.
Hosted model views can offer sideways/up-down drag, Shift+drag along the floor, wheel/pinch scaling and rotation rings. These controls affect only the authorized hosted model view. They do not grant general access to host tools or canonical world objects. Finish one mouse or touch gesture before starting another.
Moving a physics-enabled hosted model can make it static at the placed pose. Check the save result: a local saved pose does not prove that another hub or device received it. A rejected or uncertain request should be inspected before trying again. Supported local desktop and headset controls can differ from hosted browser controls.
Ask the assistant to remove a spawned object by name or recent context. If several objects match, clarify the target. Destructive-action controls still apply.
Connection interruptions
The hosted view can retain the last confirmed map, models and notes during a connection problem. Retained content is read-only until access is refreshed; it is not offline editing or an offline license. An uncertain save is not resent automatically. Use Load snapshot when a new read is needed.
A valid new snapshot can remove objects. Signing out, confirmed access denial or switching account, hub or workspace clears the private view. Reloading or closing the page may discard unsaved display changes.
Physics sandbox
Supported objects can have gravity, collisions and motion. Available activities include rigid objects, soft bodies, cloth and vehicles. Static environments and terrain depend on supported collision assets. Complex shapes may use approximate collision boundaries, so visible surfaces and contact positions can differ.
Dynamic objects remain within the supported room boundary. Static objects and some simulations have different placement limits. Object orientation and placement should remain stable across supported moves and restores; a failed operation must not be interpreted as a saved change.
Use set_physics_debug with enabled: true to inspect supported wireframes,
body bounds, joints and contacts. Set it to false to hide the overlay. Busy
scenes can omit diagnostic geometry. Ragdolls, cloth and soft bodies may have
wireframe diagnostics without filled or skinned visible meshes. Web dashboard
and terminal controls can request status or actions but do not display this
native 3D overlay. Headless operation is nonvisual.
Room navigation
bring_here recalls a supported artifact or object. Natural placement requests,
such as “Put the task board behind me,” use the available spatial controls.
Camera and headset movement remain inside the room and respect supported obstacles. Limited recovery from starting inside an object does not permit walking through arbitrary walls. Flat and XR input differ by device; Web and terminal control panels do not move their own local camera.
Mind maps
Create nodes and links to organize related ideas in space. Supported layouts spread connected groups for readability. Select a node to emphasize its connected neighbors and links.
A graph section can be appended without replacing the whole map. Named sections allow targeted updates. Supported node attributes, styles and filters can help organize a larger map; dimmed nodes remain present. Changes can use the available undo controls. Capacity and layout limits depend on the installed release.