research-document

Research Metadata Schema

Research Metadata Schema

Canonical Fields

The publisher normalizes front matter into a versioned 1.1 record with these core fields:

  • schemaVersion
  • id
  • title
  • slug
  • url
  • artifactType
  • project
  • purposes
  • audiences
  • entryPoint
  • entryPointOrder
  • entryPointLabel
  • researchArea
  • discipline
  • summary
  • status
  • version
  • confidence
  • completion
  • priority
  • authorAgent
  • created
  • updated
  • tags
  • keywords
  • relatedProjects
  • relatedDocuments
  • supersedes
  • supersededBy
  • evidenceIds
  • hypothesisIds
  • theoryIds
  • headings
  • sourcePath
  • contentHash

Legacy Aliases

The first version supports these metadata aliases:

  • identifier -> id
  • stableId -> id
  • research_area -> researchArea
  • artifact_type -> artifactType
  • updated_at -> updated
  • created_at -> created
  • author -> authorAgent
  • projectId -> project
  • purpose -> purposes
  • documentPurpose -> purposes
  • audience -> audiences

Reader-Purpose And Navigation Fields

These fields organize documents without changing their epistemic type:

  • project: stable lowercase key for the primary owning project
  • purposes: controlled reader jobs; see Document Purpose And Project Guide Architecture
  • audiences: controlled intended readers
  • entryPoint: explicit opt-in to the front of a project guide
  • entryPointOrder: numeric order within that guide
  • entryPointLabel: optional short presentation label

The build emits project entry points at data/research-guides.json. Unknown purpose or audience values produce warnings so vocabulary extensions are visible but do not silently remove content.

Compatibility Mode

When a Markdown file has no front matter, the build still publishes it in compatibility mode:

  • Title is inferred from the first H1 or filename.
  • Identifier is inferred only when a stable prefix already exists in the filename.
  • Artifact type is inferred from the directory or identifier prefix.
  • Dates fall back to the current repository implementation date.

Compatibility mode never invents canonical research claims, stable IDs, or relationship assertions.