POST
/api/memory/remember
Store a memory. Auto-linked to graph entities, embedded for vector recall. Use namespace for private agent lanes.
REQUEST
{"content": "Client prefers email follow-ups on Tuesdays",
"context": "sales call 2026-07-24",
"tier": "semantic", // short_term|episodic|semantic|procedural
"importance": 0.8, // 0-1
"agent_id": "sales-bot",
"namespace": "agent:sales-bot" // or "shared" (default)
}
RESPONSE
{"ok": true, "memory_id": 42, "linked_entities": 2,
"namespace": "agent:sales-bot", "embedding": "workers-ai/bge-base-en-v1.5"}
POST
/api/memory/recall
Retrieve memories via hybrid scoring: vector similarity (70%) + lexical (30%), then ×0.6 + recency ×0.2 + importance ×0.2. Searches your namespace + shared lane.
REQUEST
{"query": "client communication preferences",
"agent_id": "sales-bot", "namespace": "agent:sales-bot",
"top_k": 10, "tier": "semantic"}
RESPONSE
{"mode": "vector-hybrid", "results": [{"id": 42, "score": 0.81, "content": "...", "namespace": "agent:sales-bot"}]}