memory-sync

To restore agent memory continuity after a model switch, you can parse JSONL session logs to reconstruct and backfill missing daily memory files. This process recovers lost decisions, context, and relationships by analyzing historical session data.

You can detect memory gaps by running a comparison against your session logs, then backfill missing days incrementally. This syncs historical data by extracting details from JSONL files and optionally applying LLM-based summaries to create narrative memory.

Yes, you can automate nightly memory sync using cron-style scheduling. This ensures daily memory files are continuously backfilled and up-to-date by tracking local state and processing new JSONL session logs automatically overnight.

Yes, the memory backfill process includes automatic secret sanitization. This ensures that when generating LLM-based narratives from your JSONL logs, sensitive information is scrubbed before the memory files are written or summarized.

The best way to verify memory coverage is to run a gap detection command that compares existing memory files against JSONL session logs. This identifies missing days or incomplete context, allowing you to trigger an incremental backfill to restore continuity.