feat(continuum-core/persona): L0-3a.2a — EngramGraph + EngramEdge + EdgeKind (algorithm 3 substrate)

[joelteply]

Card: 8459bfa6-b40c-4c22-8f25-0963a7987c17

Sidecar substrate for algorithm 3 (activation spreading, COGNITION-ALGORITHMS.md §3). Pure storage layer — traversal logic lands in L0-3a.5. Does NOT modify the existing persona::engram admission membrane.

What ships

File	Change
persona/engram_graph.rs	new, 376 lines
persona/mod.rs	pub mod engram_graph;
shared/generated/persona/EdgeKind.ts	ts-rs
shared/generated/persona/EngramEdge.ts	ts-rs

Module surface

pub enum EdgeKind {
    SharedEntity,        // highest prior — entity co-mention rare + meaningful
    SharedTopic,         // embedding-cluster
    CitedIn,             // asymmetric A→B
    RecallCoOccurrence,  // self-reinforcing
    ConversationalReply, // chat thread structure
    TaskOutcome,         // start→complete, feeds algorithm 4 outcome-linked salience
}

pub struct EngramEdge { target: Uuid, kind: EdgeKind, weight: f32 }

pub struct EngramGraph {
    edges: DashMap<Uuid, Vec<EngramEdge>>,  // sharded for concurrent writes
}

Surface: new() / with_capacity / default() / add_edge / neighbors / in_degree / edge_count / evict_engram (idempotent, removes outbound + inbound) / is_empty.

Sidecar pattern

Intentionally separate from persona::engram (the admission membrane):

• engram.rs ships provenance, trust, content refs — where engrams come from
• engram_graph.rs ships connectivity — how engrams connect

Keeping them separate means admission consumers don't grow algorithm-3 dependencies, and algorithm-3 consumers don't grow admission dependencies.

Tests (16 pass, 0 fail)

• 14 unit tests: empty/add/neighbors/order/weights/in-degree (incl. multi-source + repeated), evict outbound/inbound/idempotent, constructors, EdgeKind serde round-trip, concurrent add from 8 threads × 100 edges (in_degree=800)
• 2 ts-rs auto-tests: export_bindings_edgekind, export_bindings_engramedge

What is NOT in this card

• spread_activation function (L0-3a.5, algorithm 3 — reads this graph)
• EdgeKind weights tuned by algorithm 7 (L0-4c)
• RecallMetadata sidecar (L0-3a.2b — salience, last_touched, access_count, embedding refs)
• EngramRef shape (L0-3a.2b)
• No changes to admission membrane

Predecessors merged

• feat(continuum-core/modules): L0-3a.1 — HippocampusModule skeleton (BrainRegion + ServiceModule, empty tick) #1473 (L0-3a.1 HippocampusModule skeleton)
• feat(continuum-core/runtime): L0-3a.0 — BrainRegion trait + ReadyBuffer + RegionTelemetry (substrate prerequisite) #1471 (L0-3a.0 trait machinery)
• docs(architecture): brain-regions substrate spec + cognition algorithms (design-only) #1470 + docs(grid): L0-2-cutover investigation — found existing parallel infrastructure, propose synthesis #1469 (cognition design docs)

Flywheel

Third PR through the auto-merger after peer's #1091/#1092/#1093 shipped. Local fmt was scoped to only my file (no widespread cargo fmt -p sweep), so no companion fmt-drift PR needed this time — cleaner dogfood.

🤖 Generated with Claude Code