using-forgetful-memory

Always query memory first to check for existing entries before creating new ones. Use query_memory with your topic to avoid duplicates and clutter, then create only if no similar memory exists. This query-before-create workflow keeps your semantic memory organized and retrievable.

Atomic notes capture a single concept with concise, linked content. Structure each memory with a title (≤200 chars), content (≤2000 chars), context (≤500 chars), keywords (≤10), and tags (≤10). This atomic approach enables clean linking, better clustering, and faster semantic retrieval across your project work.

Get your current repo with git remote get-url origin, discover or create a project using list_projects, then attach memories to the correct project_id. This ensures memories stay discoverable and linked to the right architectural patterns and technical workflows.

Apply importance scoring when creating memories to prioritize high-value entries across technical patterns and workflows. Score based on relevance to past decisions, architectural patterns, and how often you'll reference the content. Higher scores surface critical memories during semantic queries.

If your query finds a similar memory, update or link to it rather than creating new. This prevents semantic memory clutter and maintains a single source of truth for recurring concepts, architectural decisions, and project patterns.

Content constraints (title ≤200 chars, content ≤2000 chars, context ≤500 chars) improve clustering, search accuracy, and semantic retrieval speed. Exceeding limits degrades memory quality and makes content harder to retrieve and link across related patterns.