summarization

To compress scattered session memory, you extract durable decisions, constraints, and priorities from raw sources, generate structured bullet-based category summaries, and move the original files to an archive for audit.

Managing memory accumulation in long-running sessions involves identifying large memory categories flagged by relevance thresholds, extracting actionable insights, and updating the memory index to reflect newly generated summaries and archived sources.

Context synthesis prevents decision fatigue and slow retrieval by transforming fragmented memory stores into compact, durable summaries that highlight confirmed decisions, stable constraints, and current priorities for quick recall.

You should archive raw memory sources instead of keeping them active when session memory grows large, decisions accumulate, and retrieval returns too many files, ensuring future retrieval remains fast and relevant.

Yes, summarization requires reading a contexts/memory/index.md file to identify memory categories, extract durable information, and update the index to reflect new summaries and archived sources.

Yes, you can extract specific constraints and priorities from fragmented knowledge archives by identifying durable information such as validated assumptions, confirmed decisions, and known risks, then compressing them into structured summaries.