AGI MCP Server

Instructions

Implementation Reference

• src/memory-manager.js:346-377 (handler)

  Actual implementation of searchMemoriesByText method that performs full-text search on memories using PostgreSQL's tsvector and plainto_tsquery for English text search, filtering by active status and ordering by text rank and relevance score

  async searchMemoriesByText(query, limit = 10) {
    try {
      const results = await this.db
        .select({
          id: schema.memories.id,
          type: schema.memories.type,
          content: schema.memories.content,
          importance: schema.memories.importance,
          accessCount: schema.memories.accessCount,
          createdAt: schema.memories.createdAt,
          relevanceScore: schema.memories.relevanceScore,
          textRank: sql`ts_rank(to_tsvector('english', ${schema.memories.content}), plainto_tsquery('english', ${query}))`.as('text_rank')
        })
        .from(schema.memories)
        .where(
          and(
            eq(schema.memories.status, 'active'),
            sql`to_tsvector('english', ${schema.memories.content}) @@ plainto_tsquery('english', ${query})`
          )
        )
        .orderBy(
          sql`ts_rank(to_tsvector('english', ${schema.memories.content}), plainto_tsquery('english', ${query})) DESC`,
          desc(schema.memories.relevanceScore)
        )
        .limit(limit);
  
      return results;
    } catch (error) {
      console.error('Error searching memories by text:', error);
      throw error;
    }
  }

• mcp.js:554-559 (handler)

  MCP tool handler for search_memories_text that extracts query and limit arguments from the request and calls memoryManager.searchMemoriesByText()

  case "search_memories_text":
    const textResults = await memoryManager.searchMemoriesByText(
      args.query,
      args.limit || 10
    );
    return { content: [{ type: "text", text: JSON.stringify(textResults, null, 2) }] };

• mcp.js:88-104 (registration)

  Tool registration in the MCP server's ListToolsRequestSchema handler, defining the search_memories_text tool with its name, description, and input schema

  name: "search_memories_text",
  description: "Search memories by text content using full-text search",
  inputSchema: {
    type: "object",
    properties: {
      query: {
        type: "string",
        description: "Text query to search for"
      },
      limit: {
        type: "integer",
        description: "Maximum number of results",
        default: 10
      }
    },
    required: ["query"]
  }

• src/tools/memory-tools.js:61-78 (schema)

  Schema definition for search_memories_text tool, specifying the input parameters (query string required, limit integer optional with default 10)

  {
    name: "search_memories_text",
    description: "Search memories by text content using full-text search",
    inputSchema: {
      type: "object",
      properties: {
        query: {
          type: "string",
          description: "Text query to search for"
        },
        limit: {
          type: "integer",
          description: "Maximum number of results",
          default: 10
        }
      },
      required: ["query"]
    }