Summary Memory
An agent memory strategy that compresses older conversations into summaries to save tokens while retaining key long-term facts; summaries are typically generated with a small model.
Timeline (last 90 days)
The article categorizes Agent memory approaches into six types: Buffer, Window, Token, Summary, Long Term and Semantic.
Summary Memory compresses old messages into a summary before they are deleted, and the summary is usually generated with a cheap small model to control cost; in LangChain this corresponds to SummarizationMiddleware().
The three most common strategies for persisting conversation history are Window Buffer Memory, Summary Memory, and Vector Store Memory.
All Facts (3)
The article categorizes Agent memory approaches into six types: Buffer, Window, Token, Summary, Long Term and Semantic.
50%UnverifiedSummary Memory compresses old messages into a summary before they are deleted, and the summary is usually generated with a cheap small model to control cost; in LangChain this corresponds to SummarizationMiddleware().
50%UnverifiedThe three most common strategies for persisting conversation history are Window Buffer Memory, Summary Memory, and Vector Store Memory.
50%