Support multiple namespaces (subgraphs) for multi-tenant memory

[kornelrabczak]

Currently, the MCP server operates on a single flat graph, all entities and relations share the same namespace. This means one MCP server instance can only serve one use case or one user's memory at a time.

LadybugDB supports subgraphs which provide isolation similar to CREATE DATABASE / USE DATABASE in relational databases. Each subgraph has its own schema, data, and configuration.

CREATE GRAPH my_project_memory;
USE my_project_memory;
-- all operations scoped to this subgraph

Add namespace support so a single Archiledger MCP server can host multiple isolated knowledge graphs, enabling:

• Per-project memory, different projects get separate graphs
• Per-user isolation, multi-user scenarios without data leakage
• Use-case separation, memory bank vs. codebase analysis in parallel

API Design (Draft)

Add an optional namespace parameter to all MCP tools:

{
  "tool": "create_entities",
  "arguments": {
    "namespace": "project-alpha",
    "entities": [...]
  }
}

Additional management tools:

• list_namespaces, list all available subgraphs
• create_namespace, create a new isolated subgraph
• delete_namespace, drop a subgraph and all its data
• switch_namespace, set default namespace for subsequent operations