Module scheduling 

Compaction scheduling: tiered compaction dispatch, proactive compression, and background goal/subgoal extraction.

The three-tier model:

• Soft — apply deferred summaries + prune tool outputs (no LLM).
• Hard — soft tier + LLM full summarization.
• Proactive — proactively compress before the hard threshold is reached.

Functions

parse_subgoal_extraction_response

Parse the structured LLM response for subgoal extraction.