DeepFrontend
Learning Paths
Practice
System Design & DS
Design Studio
Data Structures Curriculum
Real-World Case Studies
Careers
Job Board
Resume Builder
Blog
Sign in
Search
⌘K
Go Pro
Loading compiler resources...
AI for Frontend Developers Quiz Practice
Active Recall Drill Session
← Exit Session
Question 1 of 1
hard
AI for Frontend Developers
Six months after launch, a team upgrades their embedding provider to a newer model version and updates only the query-time embedding call, leaving the vector database's stored vectors untouched. What is the correct remediation?
A
Nothing needed — newer embedding models are always backward compatible with old vectors
B
Re-ingest (re-embed and re-store) the entire knowledge base with the new model, since the old stored vectors are now in a different, incompatible vector space
C
Just increase topK to compensate
D
Switch the generation model instead
Check Answer