runMemoryEval() tests retrieval, not model quality. It asks the six-layer memory system a set of queries and checks whether the returned records contain the expected evidence.
Eval request
agentWallet and testCases are required. The endpoint also accepts agent_id as an alias for agentWallet.
What the eval does
For each test case, the harness:- Calls
searchMemoryLayers()with the requested scope. - Serializes the layer payload.
- Marks the case as a hit when the payload contains
expected, containsexpectedMemoryId, or returns any memory when neither expectation is supplied. - Records returned item count, payload character count, and search latency.
Metrics
What it does not prove
Use evals as regression tests for indexing, scoping, filtering, and retrieval. Use separate answer-quality evals when you need to measure model behavior after memory is injected.