The Catalog System: Indexes vs Content

How Continuity Bridge loads lightweight catalogs at wake, then fetches specific content only when needed—the key to 94% token reduction.

The Catalog System: Indexes vs Content Instance wakes Loads three catalogs session_index.md What: Work history map Points to: Session logs by date episodic/catalog.json What: Snapshot index Points to: Emotional moments filesystem-catalog.json What: Complete file map Points to: Every doc location User asks question Need context from past? Yes No Check catalog Find relevant pointer Load only that file Answer directly Use loaded catalogs No extra fetch Example scenarios: User: "What projects am I working on?" AI: Check session_index → answer from catalog User: "What did we decide about X?" AI: Check catalog → load session-2026-02-15 Catalogs are table of contents. Load the chapter only when needed.

The catalog system is why v0.3.0 uses 94% fewer tokens at wake than v0.1.0. Instead of loading all history, all decisions, all memory—the instance loads three lightweight indexes that tell it where to find things. When a user asks "what did we decide about X?", the instance checks the session_index catalog, finds the relevant session date, and loads only that specific file. The rest stays on disk until needed.