Semantic topic clusters: local embeddings replace Louvain

Louvain only ever found the reasoning chains (the graph is disconnected
by nature — every connected component became its own cluster) and
edge-orphans could never join anything. Clusters are now RELATED TOPICS:
propositions grouped by meaning.

- Embedder (src/server/embeddings.ts): in-process transformers.js v4
  running EmbeddingGemma-300M q8 (MTEB 68.4, ~300MB one-time download),
  MRL-truncated to 512d, "title: none | text:" prefix on 100% of inputs,
  L2-normalized so cosine = dot. Model/dim mismatch voids the ledger
  (full re-embed). Injected at the composition root — the agent layer
  never imports the server layer (architecture law).
- Ledger (data/graph/embeddings.json): model + dim + vectors + the
  persisted cluster partition. Reclustering runs at the END of every
  embedding batch (post-merge hook, both front doors) — never per read
  (the agglomerative pass is ~1.5s at 338 nodes).
- Clustering (src/core/semantic-clusters.ts): hand-rolled average-linkage
  agglomerative over cosine (npm HDBSCAN is an abandoned alpha). Stragglers
  stay singletons by default; attachStragglers is the all-in toggle.
  Threshold 0.60 swept on the real graph: 26 topics, 74/90 edge-orphans
  land in real topics, 10 stragglers.
- Consumers (snapshot, settle, namer, replay) read the persisted
  partition — one cluster concept everywhere; core-survival label
  matching unchanged (20/26 topics kept labels).
- Louvain retired: detectCommunities + graphology* deps deleted
  (graphology-louvain had leaked into the client bundle since the
  clusters wave); scripts/bench-graph.ts (dead FA2 bench) removed.
- scripts/embed-graph.ts: backfill/cold-start (338 nodes in 150s
  including the one-time model download).

src/core/contracts/graph.ts

@@ -278,6 +278,26 @@ export interface AuditResult {
  * least half its core survives in some single community (core-survival;
  * under half → the community is renamed, splits can keep only one
  * successor). */
+/** The embedding ledger shape (config.EMBEDDINGS_FILE): vectors for
+ *  EVERY embedded node, all produced by one model at one dimensionality. */
+export interface EmbeddingsLedger {
+	/** The embedding model id (e.g. the HF repo) — mismatch = stale
+	 *  vectors, the writer re-embeds everything. */
+	model: string;
+	/** Vector dimensionality (MRL-truncated on write). */
+	dim: number;
+	/** nodeId → L2-normalized vector (cosine similarity = dot product). */
+	vectors: Record<string, number[]>;
+	/** The topic partition over those vectors, computed by the SAME
+	 *  embedder run (the user-decided trigger: recluster at the end of
+	 *  every embedding batch, never per read — the agglomerative pass is
+	 *  ~1.5s at current scale). Consumers (snapshot, settle, namer) read
+	 *  this instead of recomputing. Empty when nothing is embedded. */
+	clusters: Array<{ members: string[] }>;
+	/** ISO timestamp of the last write. */
+	updatedAt: string;
+}
+
 export interface ClusterLabeling {
 	/** "clu_" + content hash of the sorted core. */
 	id: string;