Architecture

A semantic layer that crosses the four publishing fields without absorbing them.

Clearing should be implemented as a system layer, not as another publishing pipeline. Stories, Journals, Viewgraphs, and Groves continue to own canonical editorial objects. Clearing consumes explicit semantic declarations and platform-known structure from those objects, resolves a context graph, and produces bounded views of that graph for humans and machines.

The durable human-facing Clearing can be projected through a Grove because Grove already supplies identity, local navigation, pages, assets, Ground, and living editorial structure. This Grove demonstrates that choice: it is both the concept's normative reference and an example of using an existing publishing field to host a new system capability without creating a fifth field.

The Architectural Boundary

Figure 1

Rendering diagram…

The direction of ownership matters. Clearing reads canonical identity and relationships; it does not rewrite field models or move content into a new root. A Story remains a Story after it participates in a Clearing. A Journal does not need to become a Clearing document in order to challenge a thesis. A Grove can serve as the durable projection of a Clearing while remaining a Grove in the publishing architecture.

Artifact-Level Context Declarations

A future authoring contract can remain deliberately small. The illustrative shape below is a design target rather than a claim that the current runtime already accepts it.

context:
  - target: converged-enterprise
    relation: challenges

  - target: governed-data-foundry
    relation: extends

The system should resolve targets through stable Musewoods identifiers rather than brittle display titles. Validation should reject unknown targets and unsupported relationship vocabulary before the context graph is published.

A Grove used as a Clearing projection may carry additional authored orientation — canonical meaning, central question, important tensions, or Frontier — in a bounded context source or manifest extension. The implementation should prefer one semantic source of truth rather than copying those declarations into several pages for display convenience.

The Context Package

The visual Clearing should be one projection of a deeper machine-readable object. A Context Package can provide a bounded representation suitable for rendering, search, agent grounding, and future APIs.

{
  "subject": "converged-enterprise",
  "kind": "thesis",
  "status": "developing",
  "meaning": "...",
  "question": "...",
  "origins": [],
  "actors": [],
  "conditions": [],
  "tensions": [],
  "evidence": [],
  "relations": [],
  "frontier": [],
  "artifacts": []
}

The package should preserve provenance for every relationship it includes. A renderer can then group or narrate the data differently without losing the distinction between authored, reciprocal, structural, temporal, and inferred context.

Build-Time Resolution

Discover
Collect semantic declarations from canonical Musewoods objects and from the bounded source that owns a Clearing's authored orientation.
Validate
Confirm that target identifiers exist, relationship types are registered, cardinality is sensible, and prohibited contradictions or malformed cycles are caught where the system has a deterministic rule.
Resolve
Build reciprocal edges, deterministic structural relationships, chronology, and any other platform-known context while carrying provenance back to its source.
Assemble
Produce one subject-centered Context Package rather than forcing every renderer or AI caller to reconstruct the graph independently.
Project
Render a human Clearing through a Grove, generate the Context Map, expose context to search, or supply the package to an intelligent agent. Projection should not mutate the graph.

Validation Discipline

Failure	Expected behavior
Unknown target	build error with source location
Unsupported relation	build error; do not silently coerce to `related`
Duplicate edge	normalize or warn according to provenance
Impossible reciprocal rule	build error in the relationship registry
Conflicting authored claims	preserve both when genuinely contested; do not auto-resolve
Inferred edge presented as canonical	validation or projection failure
Clearing with no orientation	warning or error depending on projection contract
Frontier absent on a living subject	editorial warning, not necessarily a structural failure

The system should validate semantic impossibilities while leaving genuine intellectual disagreement visible. Two authored artifacts can disagree without the graph being invalid. Validation protects the contract; it should not pretend to adjudicate contested ideas.

Search And AI

Search can return context, not only documents
When a meaningful Clearing exists for a query, Musewoods can offer the context field as an entry surface before listing isolated matches. That moves retrieval from "which documents contain these words?" toward "what territory does this question belong to?"
Agents can enter a bounded semantic context
An intelligent agent reasoning about the Governed Data Foundry should not have to infer the standing thesis, supporting evidence, dissent, and open questions purely from vector similarity. The Context Package can supply the canonical orientation first, after which retrieval can expand into the underlying artifacts as needed.
Similarity remains subordinate to provenance
Embeddings and language models are useful for proposing candidate relationships or finding overlooked evidence. Their output belongs in the inferred provenance class until accepted by an author or another explicit governance rule. The machine may propose context; Musewoods governs meaning.