mirror of
https://github.com/docker/docker-agent.git
synced 2026-11-04 10:02:15 +00:00
When native_compaction is set, the runtime calls CompactConversation on the conversation model instead of using the LLM summarization strategy. The signed block is written to the session before the in-memory history is replaced so a crash between the two leaves a recoverable state. The post-compaction hooks fire after the block is committed. The runtime rejects native_compaction if compaction_model points at a different model than the conversation model.