Module entity 

Re-export shim over memstead_base::entity plus the git-touching git_tree_source submodule that stays in this crate.

Downstream consumers (memstead-mcp, memstead-cli, memstead-swift) reach the entity surface through memstead_git_branch::entity::* exactly as before; the inner types come from memstead-base.

Modules

generator

Entity → Markdown generator. Deterministic output for roundtrip stability.

git_tree_source

Git-tree-backed entity source — reads .md blobs out of a mem ref’s tree without a working tree on disk.

id

Entity ID parsing, generation, and path mapping.

loader

Filesystem walker — loads all .md files from mem directories into entities.

parser

Markdown → Entity parser. Handles YAML frontmatter, sections, wiki-links.

source

Entity source abstraction — where markdown comes from.

store_builder

Shared helper for turning ParseResults into a populated Store.

writer

Entity → markdown projection writer for the export paths.

Structs

Entity

A parsed entity with metadata, sections, and relationships.

EntityId

Unique entity identifier: mem--entity-path.

HeadingSpan

A single H3–H6 heading span recorded under one H2 section. Byte offsets reference the raw section content (the string stored in Entity.sections[key]). Spans are stored flat — level skips (H2 → H4 without H3) are tolerated and ancestry is resolved at query time via offset containment, not by inserting virtual intermediate spans.

ParseResult

Result of parsing a markdown file. Includes the entity, extracted inline links, and parse-time warnings (e.g. duplicate section headings).

Relationship

A declared relationship in the Relationships section.

Enums

MetadataValue

A metadata value with type coercion matching the JS parser behavior.

StubKind

Typed provenance for stub entities. Set at stub creation and lives for the engine instance’s lifetime. Boot reconstructs stubs via the parser and always tags them LoadTime; the ForwardReference and Residual variants therefore only appear during the engine lifetime in which they were created and reduce to LoadTime after a reload — this is intentional (“annotation, not state”) and consistent with the “no tombstones” decision.

Functions

normalise_description

Normalise a per-edge description from the wire/argument surface: trims surrounding whitespace and collapses empty / whitespace-only inputs to None. Applied at every mutation entry-point so the renderer never emits a bare em-dash followed by zero characters and the posture-validation step sees a canonical input.

resolve_cross_mem_refs

Rewrite relationship and inline-link targets whose slug has a "<mem>--<rest>" shape where <mem> is a visible-writable mem name. Same-mem wiki-links (no -- prefix, or a prefix that is not a known mem) are left unchanged — the resolver is additive and byte- identical for inputs without a known-mem prefix.