On this page
Operate Through a Rift๏
Reader Question๏
What does a bounded live-runtime workflow look like for a tool or agent?
Short Answer๏
Create a Rift with the required room posture, link it to an eligible frame, inspect through the viewer, bind selected objects onto the workstation, and perform allowed operations through the command surface. Codegen rooms add a validation-and-execution engine; they do not bypass the room or frame policy.
Representative Shape๏
# After Nexus configuration and activation for this process:
rift = nexus.create_rift(rift_name="operations")
root = rift.create_nexus_frame(frame_name="ops")
rift.create_frame_link("ops")
space = rift.space
viewer = space.frame_viewer
workstation = space.workstation
command = space.command_system
service = root.meld("ReportService")
workstation.bind_object("service", service, weak_ref=False)
workstation.set_target("service")
The viewer and command system continue to resolve current Rift projection truth. The workstation is a room-local workbench, not a second ownership system for Melder's world.
Why This Design Is Strong๏
The operator receives a named place with a bounded vocabulary.
Selected live objects can be worked on without making the whole process globally visible.
Static, capability, and codegen needs do not share one maximal authority surface.
Tradeoffs๏
The workflow is more explicit than passing a service object directly to a tool. That explicitness creates an auditable boundary for discovery, targeting, action, and refresh.
Where to Go Next๏
Evidence: