SuperbeeDocs
v0.1.3Source repository

Concept

Understand reusable domain structure

Choose when documents, relationships, Kinds, recipes, and Views earn their maintenance cost.

View Markdown

Question answered

When should a bundle use an ordinary document, Kind, recipe, relationship, or View?

This explanation is for active users and agents choosing how much durable structure a real problem needs. It is grounded in the current stable release and the current Superbee Skill.

Start with the recurring cost

Structure earns its place when it prevents repeated reconstruction, inconsistent records, lost provenance, or repeated human interpretation. Begin with the smallest layer that removes the observed cost. Add another layer after use reveals a stable need.

LayerUse it whenWhat it addsOngoing cost
DocumentOne important fact, decision, result, or explanation needs to persist.A typed Markdown record with a stable ID.Keep one record accurate.
RelationshipThe meaning between two records matters to later readers or queries.A labeled Markdown link and a derived backlink.Preserve accurate targets and vocabulary.
KindSeveral documents represent the same domain concept and consistency matters.Shared fields, allowed values, headings, paths, freshness, and relationship guidance.Migrate instances when the convention evolves.
RecipeStable definitions should be installed in another bundle.A portable package of conventions and declared reference or View definitions.Version and review the reusable package.
ViewPeople repeatedly rebuild the same comparison, graph, timeline, or decision surface.A human presentation over authoritative bundle knowledge.Maintain the presentation and its trust boundary.

How the layers fit together

Documents preserve specific knowledge

A document records one meaningful thing in ordinary Markdown with YAML frontmatter. Generic documents remain useful throughout a bundle's life because many facts never recur enough to justify a schema.

Create a representative document while a domain shape is still emerging. Its actual use reveals which fields, headings, and relationships remain stable.

Relationships preserve context between records

Links can express evidence, dependency, provenance, containment, or any vocabulary the domain needs. Superbee stores the link in the source document and derives the backlink. This keeps the graph inspectable in Markdown and avoids a separate edge database.

A Kind can declare typed relationship vocabulary and expected inbound relationships. Untyped citations remain available for links whose meaning has not stabilized.

Kinds make recurring concepts consistent

A Kind is a convention document owned by the bundle. It governs documents whose type matches its governs value. The convention may declare fields, enum values, required headings, path prefixes, freshness horizons, terminal workflow values, and relationship vocabulary.

Kinds support reliable creation and maintenance. superbee new "<Kind>" validates a new instance strictly, while superbee status checks existing instances across the bundle. The documents remain the records people and agents read.

Recipes move stable definitions between bundles

A recipe contains definitions and guidance. Applying one copies its declared convention documents into the selected bundle, which then owns those files. Project instances stay with their project.

Recipe application is idempotent. If an installed definition has changed locally, a later recipe application reports source drift and leaves the bundle's version untouched. This behavior gives the owner a visible migration decision.

Views reduce repeated interpretation

A View is useful when a human repeatedly needs the same representation of several documents and relationships. Examples include a research evidence map, release readiness panel, experiment comparison, or interactive timeline.

The View reads admitted bundle knowledge through an explicit capability. Documents and evidence remain authoritative. The View can evolve or be replaced without migrating the underlying domain records.

A typical progression

Consider a team learning from customer interviews:

  1. Preserve the first interview as an ordinary Interview document.

  2. Add links when statements support existing decisions or reveal related needs.

  3. After several interviews show the same useful fields and sections, declare an Interview Kind.

  4. Introduce an Insight Kind after repeated synthesis exposes a stable shape and relationship vocabulary.

  5. Package the definitions as a recipe when another bundle needs the same reviewed model.

  6. Add a View after stakeholders repeatedly assemble the same evidence and insight overview.

Each step answers a cost observed in use. A project may stop at any step. A research bundle can benefit from Kinds and relationships without adopting task tracking or a roadmap.

Selection check

Use these questions before adding structure:

  1. Is there one important record or a recurring class of records?

  2. Which inconsistencies have already caused work or confusion?

  3. Do links carry useful domain meaning that later readers need to query?

  4. Will the same definitions be reused in another bundle?

  5. Which overview do people repeatedly reconstruct?

Choose a document for the first case, a Kind for proven repetition, relationships for durable context, a recipe for reusable definitions, and a View for repeated interpretation. If the evidence is still weak, keep a representative document and revisit the decision after use.

Boundaries and limitations

  • A Kind validates declared shape. It cannot determine whether a claim, decision, or conclusion is true.

  • A recipe installs definitions. It does not manage project instance data.

  • A View presents bundle knowledge. It does not become the authority for that knowledge.

  • Convention changes can create conformance work for existing documents. Inspect the affected instances and plan the migration before changing the convention.

  • Different purposes may deserve separate bundles when their privacy, participants, or lifecycle differ.

Evidence

The released implementation and tests for these boundaries are:

  • Kind registry and vocabulary

  • Recipe parser

  • Typed relationship tests

  • Registered View implementation