Specification Guide

The normative specification lives in docs/spec. The repository now organizes the ARC RDM model as three sibling profiles: ARC Core, datamap, and administrative. The implementation remains one shared model; profile boundaries are documentation and mapping boundaries, not runtime type boundaries.

Unified Profiles

flowchart LR
    Dataset --processes--> Process
    Dataset --dataFiles--> Data
    Dataset --agents--> Agent
    Dataset --citations--> ScholarlyArticle
    Dataset --dataContexts--> DataContext
    Dataset --hasPart--> Dataset
    Process --inputs--> Sample
    Process --"outputs"--> Data
    Process --executesRecipe--> Recipe
    Process --parameterValue--> Annotation
    Recipe --parameters--> FormalParameter
    Recipe --components--> Annotation
    Annotation --instanceOf--> FormalParameter
    Agent --affiliation--> Organization
    ScholarlyArticle --authors--> Agent

Shared model entities:

Entity	Source
Dataset	Dataset
Process	Process
Recipe	Recipe
Sample	Sample
Data	Data
DataContext	DataContext
Annotation	Annotation
FormalParameter	FormalParameter
DefinedTerm	DefinedTerm
Agent	Agent
Organization	Organization
ScholarlyArticle	ScholarlyArticle

Profile entry points:

Profile	Purpose	Source
ARC Core	Provenance through datasets, processes, protocols, samples, data, and annotations	ARC Core
Datamap	Data files, data fragments, fragment descriptors, and dataset data contexts	Datamap
Administrative	Dataset metadata, agents, affiliations, citations, licenses, and dates	Administrative

Decorations

Decorations add domain-specific meaning through additionalType, specialized properties, and decoration-specific entities.

Decoration	Purpose	Source
ISA	Investigation, Study, Assay, Source, Sample, and ISA property value roles	ISA
Workflow Run	Workflow and Run datasets, workflow protocols, and workflow invocations	Workflow Run
Datamap	Promoted to a sibling profile	Datamap

Naming Notes

The current core vocabulary uses Process and Recipe, not the older placeholder names Process and Protocol.

For process I/O, the current core and YAML schema names are:

inputs
outputs
executesRecipe
parameterValue
additionalProperty

Some legacy/profile-shaped examples and upstream references use RO-Crate or Bioschemas names such as object and result. See Examples and schemas for how those files are treated.

Querying

The query use cases are described in Querying. The implementation exposes Path as a returned value object, while traversal and query operations are attached to the model types in src/ProcessCore/Graph.fs.