create_memory_cluster

Instructions

Implementation Reference

• src/memory-manager.js:217-238 (handler)

  The core handler implementation that creates a new memory cluster in the database. Accepts name, clusterType, description, and keywords parameters, creates a default centroid embedding, and inserts the cluster into the memoryClusters table.

  async createMemoryCluster(name, clusterType, description, keywords = []) {
    try {
      const defaultEmbedding = new Array(1536).fill(0.0);
      
      const [cluster] = await this.db
        .insert(schema.memoryClusters)
        .values({
          name,
          clusterType,
          description,
          keywords,
          centroidEmbedding: defaultEmbedding,
          importanceScore: 0.0
        })
        .returning();
  
      return cluster;
    } catch (error) {
      console.error('Error creating memory cluster:', error);
      throw error;
    }
  }

• src/tools/memory-tools.js:127-155 (schema)

  Tool schema definition for create_memory_cluster in the memory tools module. Defines input parameters (name, cluster_type, description, keywords) with cluster_type being an enum of valid cluster types.

  {
    name: "create_memory_cluster",
    description: "Create a new memory cluster",
    inputSchema: {
      type: "object",
      properties: {
        name: {
          type: "string",
          description: "Name of the cluster"
        },
        cluster_type: {
          type: "string",
          enum: ["theme", "emotion", "temporal", "person", "pattern", "mixed"],
          description: "Type of cluster"
        },
        description: {
          type: "string",
          description: "Description of the cluster"
        },
        keywords: {
          type: "array",
          items: { type: "string" },
          description: "Keywords associated with this cluster",
          default: []
        }
      },
      required: ["name", "cluster_type"]
    }
  },

• mcp.js:576-583 (handler)

  MCP handler routing that receives tool calls, extracts arguments (name, cluster_type, description, keywords), calls the memoryManager.createMemoryCluster method, and returns the result as JSON.

  case "create_memory_cluster":
    const newCluster = await memoryManager.createMemoryCluster(
      args.name,
      args.cluster_type,
      args.description,
      args.keywords || []
    );
    return { content: [{ type: "text", text: JSON.stringify(newCluster, null, 2) }] };

• mcp.js:153-180 (registration)

  Tool registration in the MCP server's tools list. Defines the tool name, description, and input schema that is exposed to MCP clients.

  {
    name: "create_memory_cluster",
    description: "Create a new memory cluster",
    inputSchema: {
      type: "object",
      properties: {
        name: {
          type: "string",
          description: "Name of the cluster"
        },
        cluster_type: {
          type: "string",
          enum: ["theme", "emotion", "temporal", "person", "pattern", "mixed"],
          description: "Type of cluster"
        },
        description: {
          type: "string",
          description: "Description of the cluster"
        },
        keywords: {
          type: "array",
          items: { type: "string" },
          description: "Keywords associated with this cluster",
          default: []
        }
      },
      required: ["name", "cluster_type"]
    }

• drizzle/schema.ts:37-56 (schema)

  Database schema definition for the memoryClusters table including fields for id, clusterType, name, description, centroidEmbedding, keywords, importanceScore, and various indexes for performance optimization.

  export const memoryClusters = pgTable("memory_clusters", {
  	id: uuid().defaultRandom().primaryKey().notNull(),
  	createdAt: timestamp("created_at", { withTimezone: true, mode: 'string' }).default(sql`CURRENT_TIMESTAMP`),
  	updatedAt: timestamp("updated_at", { withTimezone: true, mode: 'string' }).default(sql`CURRENT_TIMESTAMP`),
  	clusterType: clusterType("cluster_type").notNull(),
  	name: text().notNull(),
  	description: text(),
  	centroidEmbedding: vector("centroid_embedding", { dimensions: 1536 }),
  	emotionalSignature: jsonb("emotional_signature"),
  	keywords: text().array(),
  	importanceScore: doublePrecision("importance_score").default(0),
  	coherenceScore: doublePrecision("coherence_score"),
  	lastActivated: timestamp("last_activated", { withTimezone: true, mode: 'string' }),
  	activationCount: integer("activation_count").default(0),
  	worldviewAlignment: doublePrecision("worldview_alignment"),
  }, (table) => [
  	index("memory_clusters_centroid_embedding_idx").using("ivfflat", table.centroidEmbedding.asc().nullsLast().op("vector_cosine_ops")),
  	index("memory_clusters_cluster_type_importance_score_idx").using("btree", table.clusterType.asc().nullsLast().op("enum_ops"), table.importanceScore.desc().nullsFirst().op("float8_ops")),
  	index("memory_clusters_last_activated_idx").using("btree", table.lastActivated.desc().nullsFirst().op("timestamptz_ops")),
  ]);