[doc] Support TTL on ShortTermMemory in User Documentation

[da-daken]

Linked issue: #753 #580

Purpose of change

Add a document for Support TTL on Short Term Memory

Tests

No need to test

API

no api

Documentation

• doc-needed
• doc-not-needed
• doc-included

[da-daken]

hi @xintongsong @GreatEugenius , I have added user documentation and configuration documentation about short-term memory TTL, but I'm not sure if it meets the requirements. PTAL～

[weiqingy]

Thanks for adding this — nice that the Python and Java snippets mirror the real e2e/integration tests exactly, and every config key, default, and enum value matches AgentExecutionOptions / core_options.py. One optional question inline.

[weiqingy]

Both the prose ("older state expires automatically… expiring historical data") and the two examples use ON_READ_AND_WRITE, the default — under which a read also resets the TTL clock. So an old entry that's still being read periodically won't expire. A reader skimming the intro might take "older" to mean by creation time. Would it be worth one clarifying sentence here — that with ON_READ_AND_WRITE "age" is measured from last access, and ON_CREATE_AND_WRITE is the one that expires by write/creation time? It's okay with me if you'd rather keep it brief.

[da-daken]

I think 'controls whether TTL is refreshed on create/write or on read/write.' already sufficiently explains the above situation, so I don't intend to provide any additional supplements. Thank you for such a prompt reply.

[wenjin272]

I believe the literals themselves, along with a single line of additional explanation, are sufficient; therefore, I agree with keeping it brief.

[wenjin272]

LGTM