Description
Currently, the DreamService.runGarbageCollection() pipeline successfully decays topological edges (relationships) to maintain structural integrity. However, the exact underlying vector data (the nodes) inside the SQLite engine remains permanently stored even when isolated.
We need to implement "Memory Apoptosis" (cellular cleanup). When a node (such as an old CONCEPT or EPISODE) decays to the point where it loses all inbound and outbound structural edges (weight 0), the runGarbageCollection phase should permanently delete the underlying Vector embedding from the .neo-ai-data/neo.db.
Objective
- Expand
GraphService or SQLiteVectorManager with a hard-delete API (deleteNode).
- Enhance
DreamService garbage collection to query for orphaned nodes post-edge-decay and delete them.
- Ensure orphaned nodes are completely wiped from both structural tables and vector embeddings to keep semantic search extremely fast and hallucination-free.
Description
Currently, the
DreamService.runGarbageCollection()pipeline successfully decays topological edges (relationships) to maintain structural integrity. However, the exact underlying vector data (the nodes) inside the SQLite engine remains permanently stored even when isolated.We need to implement "Memory Apoptosis" (cellular cleanup). When a node (such as an old
CONCEPTorEPISODE) decays to the point where it loses all inbound and outbound structural edges (weight 0), therunGarbageCollectionphase should permanently delete the underlying Vector embedding from the.neo-ai-data/neo.db.Objective
GraphServiceorSQLiteVectorManagerwith a hard-delete API (deleteNode).DreamServicegarbage collection to query for orphaned nodes post-edge-decay and delete them.