SQL & Vector-Adjacent Data
Relational SQL for users, billing, and strict filters; optional pgvector in the same engine for semantic recall — plus the Prisma-shaped mental model used across this stack.
Step 1 · concept
Beyond the Vector
While Vector Databases are "hot," 90% of your application's logic will still live in a traditional relational database (SQL).
You need SQL for:
- User State: Who is the learner? What is their progress?
- Conversation Metadata: Which lesson does this chat belong to?
- Rate Limiting: How many messages has this user sent this month?
The "modern" approach is to use a database that does both, like Postgres with pgvector.
ACID