Database
PostgreSQL via Prisma
3ms
pgvector extension
Required for embedding similarity search
enabled
LLM scoring
Model: google/gemini-2.5-flash-lite
enabled
Embeddings
all-MiniLM-L6-v2 · 384-dim · runs in-process
local (ONNX)
Typical search time
Cold search · warm repeat searches are faster
60-70 s
Raw JSON: /api/health