| name | kg-memory |
| skill_type | skill |
| description | Thin verb over the engine's EG-KG.memory.eg-batch-decay-caller memory surface — episodic→semantic consolidation, the spatial scene graph, RL trajectories, plus unified store/recall/link memory-CRUD. Use for agent memory ops — "consolidate memory", "store/recall a memory", "append a trajectory step", "scene transform". |
| license | MIT |
| tags | ["graph-os","engine","memory"] |
| tier | core |
| metadata | {"author":"Genius","version":"0.1.0"} |
kg-memory
Condensed intent-surface note (Seam 8). Under the default intent surface (MCP_TOOL_MODE=intent), graph_memory is held back from the default tool list (nothing removed — REST + _execute_tool still reach it exactly as documented below). Two ways to use this skill unchanged: (1) load_tools(tools=["graph_memory"]) once per session (as below), then proceed exactly as documented; or (2) call the write intent verb with the same natural-language request — the resolver routes to graph_memory for you and returns the result plus a routing justification. Set MCP_TOOL_MODE=condensed/verbose/both to expose the granular tools eagerly instead.
graph_memory (CONCEPT:AU-KG.coordination.engine-message-broker) fronts the engine's memory surface. Engine methods (action-routed 1:1, dashes→underscores): create_summary, consolidate, maintain, add_scene_object, world_transform, start_trajectory, append_step, discounted_return, get_*. Unified memory-CRUD (AU-KG.memory.unified-memory-crud-core) store (agent_id+content[+memory_type,tags]), recall (query[+memory_type]), link (source+target[+rel_type]) route into the SAME graph_write memory core as the REST /graph/write/memory twins. Structured args go via params_json.
Invoke
- MCP:
load_tools(tools=["graph_memory"]), then graph_memory(action="recall", params_json='{"query":"..."}').
- REST twin:
POST /graph/memory with {"action": "consolidate"}.
Example
graph_memory(action="store", params_json='{"agent_id":"a1","content":"...","memory_type":"semantic"}')
