Which approach keeps embeddings synchronised when source text is updated?
- A A trigger or application-side pipeline that recomputes the embedding on UPDATE (correct answer)
- B A foreign key constraint
- C A unique index on the vector column
- D Setting the vector column to GENERATED
Shortcut: Embeddings are derived data. A trigger enqueuing a re-embed job, or explicit application logic on write, keeps the vector consistent with the text it represents.