para-memory-files

To organize persistent knowledge across sessions using the PARA method, this system stores durable facts in a structured knowledge graph under life/ folders, captures raw timelines in daily notes, and applies memory-decay rules during weekly synthesis. It separates entities, daily logs, and tacit patterns into three distinct layers.

The best way to manage daily notes for knowledge management is to save raw timelines as memory/YYYY-MM-DD.md files, then extract durable facts into the knowledge graph's items.yaml during heartbeats. This ensures temporary logs are processed into structured entity summaries without manual duplication.

Yes, you can track tacit user patterns and preferences over time using a dedicated MEMORY.md file. This third memory layer captures long-term lessons and behavioral patterns, updating gradually alongside the knowledge graph and daily notes to maintain a persistent user profile.

Weekly synthesis works by reviewing accumulated daily notes and knowledge graph entities, applying memory-decay rules to archive outdated information, and consolidating remaining context. This process triggers via qmd queries to recall past context and refresh planning files without deleting historical data.

Yes, you need YAML files to store atomic facts when building this knowledge graph. Each entity folder requires a summary.md for overviews and an items.yaml file to structure durable facts, manage fact-status, and link related entities within the PARA folders.

You query past context and manage plans by triggering qmd for semantic search recall and storing plans under a shared plans/ directory for cross-agent access. This approach ensures planning files remain separate from the core memory layers while remaining accessible.