[KongchangAI]
ConceptToken记忆

Token Memory

An agent memory strategy that trims historical messages according to a token budget, discarding the portion exceeding the budget, with boundary handling performed by trim_messages().

Timeline (last 90 days)

Oct 18

The article categorizes Agent memory approaches into six types: Buffer, Window, Token, Summary, Long Term and Semantic.

Unverified50%
Oct 18

Token Memory uses token count rather than message count as the budget cap, and trimming may cut a sentence off in the middle, leaving the semantics incomplete.

Unverified50%

All Facts (2)

Source Articles