feat(memory): hybrid vector search via embedder

- repository: optional pgvector embedding column (migration with
  CREATE EXTENSION guarded by main), SearchByVector cosine search,
  vector-enabled insert path
- service: SetEmbedder — memories are embedded on Add; Search merges
  vector hits with n-gram keyword matches (vector first, deduped)
- embedder was already implemented (OpenAI-compatible API + n-gram FNV
  fallback) but never instantiated; main now wires it from
  EMBEDDING_API_URL/EMBEDDING_MODEL config

backend/cmd/cyrene/main.go

@@ -133,6 +133,16 @@ func main() {
 	if database != nil {
 		memRepo := memory.NewRepository(database)
 		memSvc = memory.NewService(memRepo)
+
+		// Embedding: real API when configured, n-gram FNV fallback otherwise
+		emb := memory.NewEmbedder(cfg.EmbeddingAPIURL, cfg.EmbeddingAPIKey, cfg.EmbeddingModel)
+		if err := db.RunMigration(context.Background(), database, "memories-vector", db.SplitSQL(memory.VectorMigrationSQL(emb.Dims()))); err != nil {
+			logger.Infof("[memory] pgvector unavailable, keyword-only search: %v", err)
+		} else {
+			memRepo.EnableVector()
+			memSvc.SetEmbedder(emb)
+			logger.Infof("[memory] hybrid search enabled (vector dims=%d, api=%v)", emb.Dims(), emb.IsAvailable())
+		}
 		logger.Infof("[memory] connected")
 	} else {
 		logger.Infof("[memory] disabled (no database)")