ai lifecycle / embeddings
Embedding pipeline
Every vector table on this instance, with live embedding counts straight from the engine's vector registry. Vectors index into HNSW on write — no separate ingestion job to babysit.
instance ·
vector tables
—
total embeddings
—
index
HNSW · int8
ingest queue
—
counts are live from the vector registry · per-second ingest-rate timeline lands with the metrics pipeline