Skip to content

Model

The developer's by-kind index: where each CORA aggregate's MOGNO content lives, the new Sirius Site and the compute axis named for reconstruction, and the record of what is deliberately deferred. Design-phase scaffold.

For the aggregate shapes see the architecture model and the per-BC modules.

Aggregate (BC) Where at MOGNO
Asset (Equipment) the stage pages: Source, Sample, Detector
Computed / virtual axes (Equipment) Source (the cone-beam magnification PseudoAxis)
Capability, Method (Recipe) Techniques
Enclosure (Enclosure) the index
Zone, Conduit, Policy (Trust); Actor (Access) Governance
Procedure, Recipe, Caution, Supply, Subject, Run, Campaign, Dataset, Decision deferred (design-phase; see below)

What makes MOGNO new

MOGNO is CORA's eighth Site (Sirius, the Brazilian Synchrotron Light Laboratory at CNPEM) and the fleet's first South American facility, the biggest re-test of the Site and Federation kernel a single deployment can be. Its science is cone-beam X-ray micro and nanotomography across two endstations, fed by a quasi-monochromatic dipole source at three working energies.

It is also the thinnest reverse-engineered scaffold to date, by necessity. The NSLS-II and Diamond beamlines were built from public open-source controls libraries (bluesky profile collections, dodal), so their descriptors carry real EPICS PVs or device handles. MOGNO has no public controls configuration: its facts come from two papers and a facility page. So every device binds a catalog Family but carries no handle and no vendor Model; the handles are open questions, not read-from-config evidence.

No new families, no new method (reuse and reinforce)

MOGNO coins nothing. It is the tomography spine landing on a third facility after the APS 2-BM pilot and NSLS-II FXI, and it reuses their vocabulary wholesale:

  • the rotation axes bind RotaryStage (the master clock for triggered acquisition);
  • the sample positioners (including the fine piezo "tripod") bind LinearStage, the axis set a per-Asset setting;
  • the focusing optics bind Mirror, the beam-defining slits bind Slit;
  • the detectors bind Camera and Scintillator (the indirect chain);
  • the cone-beam magnification binds PseudoAxis (the FXI Magnification precedent);
  • the TATU FPGA trigger binds TimingController (the 2-BM softGlueZynq / FXI Zebra precedent);
  • the machine state binds the loose StorageRing;
  • the tomography Method is reused, the practice carried pending (TECH-1).

The control seam: a custom EPICS application layer

MOGNO is the fleet's first orchestration layer that is neither Bluesky nor BLISS nor Sardana. Its floor is EPICS IOCs plus a TATU FPGA trigger/timer (shared Sirius infrastructure, exposing EPICS PVs via the LNLS Nheengatu layer). Above that floor sits a beamline-owned custom PyEpics application stack: mgn-devices (device abstraction over PyEpics), mgn-routines (the alignment and tomogram acquisition routines), and mgn-control-guis (the PyQt/PyDM launchers). A tomogram is launched from a GUI dialog that runs the relevant mgn-routines script as a subprocess, driving the rotation stage, detector, and TATU trigger over EPICS.

That custom routine layer is the orchestration CORA's edge would conduct over the EPICS floor, exactly as the 2-BM seam designates: CORA's EdgeConductor replaces the scan/alignment orchestration the mgn-routines perform today, conducting over the EPICS + TATU floor rather than replacing it. The beamline has named Bluesky/sophys (Ophyd devices, Bluesky plans) as a future migration target; whether it has migrated is an open question (ORCH-1). Modelling the seam against a custom stack, not Bluesky, is the point: it confirms the seam model does not assume a particular orchestration framework.

The compute axis: reconstruction named, not built

MOGNO reconstructs on an HPC cluster. The beamline's reconstruction library is ssc-raft (CUDA, from the Sirius Scientific Computing group), submitted to the cluster over SSH from a FastAPI reconstruction service with a PyQt job-queue GUI; it has been the production reconstruction path since early 2024. This is a clean instance of CORA's compute axis (a ComputePort over a Method, no new BC): named here as reinforcement of the compute-modelling synthesis, not modelled as Assets in this cut. The cluster name, scheduler, and storage path are not confirmed from public sources (COMPUTE-1).

Deliberately not here yet

  • The exact device handles and vendor models (CTRL-1, STAGE-1, STAGE-2, STAGE-3, CAM-1, CAM-2, OPT-1, OPT-2). No public controls config exists; the PVs, controller boxes, and part numbers must come from staff. They carry no value here rather than a guessed one.
  • The reconstruction Assets and the compute leg (COMPUTE-1). The ssc-raft HPC path is named above, not modelled.
  • The orchestration migration status (ORCH-1). Whether MOGNO still runs the custom mgn-* stack or has moved to Bluesky/sophys is an open question.
  • The PSS permit signals and shutters (PSS-1). Absent from public sources, carried pending, not invented.
  • The simulated devices and full asset-tree scenarios. No test_mogno_*.py registers the asset tree, and no vendor Models are bound.
  • Operations and experiment views. A runbook and live experiment view for a beamline CORA does not yet drive would be invention; see the note on the index.