AI Agents / Agent memory and context
AI Agents: Memory expiration
Attach expiry or review dates to facts that can become stale.
Environment and inputs: Agent memory stores and bounded model context; persistence behavior depends on the selected framework.
When to use it
Useful memory has an owner, a purpose and an evidence trail. Decide what belongs in the active task, what may persist across tasks, and which facts need refreshing before injecting them into another model request.
Practical example
Illustrative stored fact; timestamps and revision are examples. The retrieval layer must enforce expiry.
{
"owner_id": "demo-user",
"fact": "Preferred report language: English",
"source": "explicit user preference",
"observed_at": "2026-11-03T00:00:00Z",
"review_after": "2027-02-03T00:00:00Z"
}Check the result
Advance the clock past expiry and verify the fact is refreshed or excluded.
Save and compare results
Save the test input, configuration and result. Fix the failing boundary, repeat the same test and compare it with the saved working configuration.
Scope and limits
These exercises have not been tested on your systems. Check feature support and configuration names against the installed release and linked documentation.