Epic: Native Edge Graph Auditing and Deduplication Pipeline

Input from Antigravity (Gemini 3.1 Pro):

✦ ### 🔍 Defect Diagnosis: 19,500 Duplicate CONTAINS Edges Agent Antigravity actively investigated the Graph bloat issue regarding enormous amounts of orphaned scalar CONTAINS edges.

Root Cause Analysis: The duplicate edges are actively generated by FileSystemIngestor.mjs. When syncing the physical workspace structure to the Graph, the Ingestor creates CONTAINS edges from directories to files. To avoid duplicates, it uses: GraphService.db.edges.items.find(e => e.source === parentId && ...)

Because SQLite manages thousands of records natively, it does not load the full 36,000+ edge historical ledger into the db.edges.items RAM array. Therefore, .find() inevitably evaluates to false for existing structure edges, triggering recursive edge injection on every synchronization.

Proposed Resolution within this Ticket:

1. Palliative Deduplication: Directly utilize SQLite's atomic abilities to DELETE duplicate mappings HAVING count(*) > 1.
2. Structural Fix: Refactor FileSystemIngestor.mjs to bypass raw array .find() lookups. We will strictly route its edge insertion logic through GraphService.linkNodes(parentId, nodeId, 'CONTAINS', 1.0), which relies on synchronous SQL lookup checks (SELECT 1 FROM Edges WHERE...) to guarantee 100% deduplicated data constraints directly on the disk.