Memory Distillation

A context management strategy for large language models in which messages that exceed the context window limit are compressed into a summary and retained. This approach preserves the semantic content of historical information within the constraints of a limited context window.