Skip to content

Capabilities

Recipe BC Capabilities. A Capability is the operations-layer template that declares WHAT an operation provides: the Affordances any wired Asset's Families must cover (required_affordances), the parameter contract any binding Method must be a structural subset of (parameter_schema), and which executor kinds may bind (executor_shapes — closed v1 enum {METHOD, PROCEDURE}). Methods bind to one Capability via capability_id; at define_plan time, every wired Asset's Families' affordances must cover the bound Method's Capability's required_affordances (otherwise PlanAffordancesNotSatisfiedError, 409). Capabilities are cross-facility vocabulary: codes are namespaced cora.capability.* (closed core) with cora.capability.<facility>.* available for site-specific extensions. See Model for the aggregate shape.

Closed core (v1)

Code Name Binds Methods Notes
cora.capability.tomography Tomography tomography, streaming_tomography, continuous_rotation_tomography, mosaic_tomography The canonical synchrotron-CT science technique. Direct analogue to NeXus NXtomo / PaNET tomography. Executor shape: METHOD.
cora.capability.acquisition Acquisition first_light, dark_baseline, flat_baseline The operational primitive for frame-stack capture. Distinct from tomography — acquisition is the bare "produce frames" primitive (NeXus NXscan analogue); tomography is the specific rotational-projection technique on top.
cora.capability.alignment Alignment resolution_alignment, focus_alignment, center_alignment, roll_alignment, pitch_alignment Iterative tuning toward a target metric. Each Method is a distinct step in the rotation-axis alignment chain; the Capability is the shared operational primitive (iterate-observe-adjust).
cora.capability.calibration Calibration alignment_calibration Empirical measurement of system constants (motor sensitivities, beam profiles, encoder offsets). Distinct from alignment — calibration measures constants, alignment consumes them.
cora.capability.maintenance Maintenance motor_homing, hexapod_reboot One-shot setup, recovery, and ceremony operations with no scientific Dataset output. Currently exercised via both METHOD (recipe-ladder anchoring) and PROCEDURE (ISA-106 step-by-step ceremony executor) shapes.

Naming conventions

  • cora.capability.<snake_case> — lowercase, regex ^[a-zA-Z0-9_-]{1,64}$ (matches MCP / OpenAI / Anthropic tool-name constraints so the unprefixed tail can be exposed verbatim on agent tool surfaces).
  • Noun or gerund only, never verb. Verbs belong at the Affordance (-able adjective) or agent-tool layer (get_*, read_*). Capability names describe what an operation provides.
  • One word when a community shorthand exists (tomography, alignment, acquisition); head-noun-last snake_case compound only when one word is genuinely ambiguous (e.g., future energy_scan vs ambiguous bare energy).
  • Artifacts and milestones are NOT Capabilities. baseline is a Dataset kind; first_light is a milestone (the Method captures the operational acquisition, the milestone-name stays only because the synchrotron community universally calls it that). Data-reduction steps (e.g., flat_field_correction) live in external pipelines, not as CORA Methods.

Catalog governance

  • Closed core under cora.capability.*, with cora.capability.<facility>.* available for site-specific extensions on demand (FHIR/LOINC status + replaced_by precedent).
  • Status FSM Defined → Versioned → Deprecated; deprecated codes carry optional replaced_by_capability_id for cross-version lineage.
  • executor_shapes is REQUIRED non-empty at definition. Closed v1 set: {METHOD, PROCEDURE}. A Capability may declare both.
  • Adding a value is additive only — deprecate-and-replace, never remove. Replay safety.

Pending in code

  • cora.capability.energy_scan — Method-binding for energy_change style scans (currently energy_change binds to tomography).
  • cora.capability.steering — Method-binding for closed-loop autonomous-CT adjust_run flows.
  • Methods — the Method catalog, with each Method's Capability binding in the second column.
  • Affordances — the closed 28-item primitive set Capabilities declare as required_affordances.
  • Glossary — Capability vs Family vs Affordance vs Method distinctions.