KEVP — Kemet Embedding Vector Protocol

1. Charter

KEVP defines how public posts and Node interest profiles embed into a shared semantic index. Embeddings power KCAP, KDRP, and KSRP without exposing raw model weights or serving topology.

2. Ingest Lifecycle

1. Post created → language routed by KLDP.
2. Policy-eligible text/media → embedding computed at edge ingest.
3. Vector written to semantic index with post identity and freshness markers.
4. Policy removal or deletion triggers vector retirement from active retrieval.

3. Interest Profiles

Viewer interest vectors aggregate engagement over time with decay. Cold or missing vectors delegate to KCAP-CS priors. We do not publish aggregation coefficients.