The Graph IS the Vector Space

In every other system, the index and the data are separate structures that must be kept in sync. SERAPH has one structure. A frame's semantic parent is its nearest neighbor in embedding space. Eigenframes — the high-traffic hubs that become fast-access entry points — aren't designated by a curator. They emerge from topology because many other frames selected them as parents. The graph is what the vector space looks like when you read it as a graph.

This means:

• No index synchronization. There is no second structure to corrupt.
• No stale-index bugs. The bug class doesn't exist.
• Full reconstructability. If metadata is lost, replay the frame log. The graph rebuilds identically because the graph is determined by the frames.

	Vector DB	Knowledge Graph	SERAPH
Topology	None. Flat index.	Authored schema.	Emergent from semantic proximity at ingest.
Relationships	External join required.	Manually extracted triples.	Emergent edges from embedding proximity, plus explicit typed edges — both in one store.
Structure cost	Zero (no structure).	High (extraction pipeline).	Emergent structure is automatic. Explicit structure is opt-in, zero-schema.
Grows with data	Index grows. No structure grows.	Only with manual extraction.	Every frame adds edges. Typed edges add curated structure. Both enrich the same topology.