Section checkpoint

Embeddings, Vector Search & RAG

12 questions drawn from the 6 chapters in this section. If most land on the first try, you are ready for the next section. If not, the explanation after each answer says which idea to revisit.

Checkpoint

12 questions · first-try answers earn the most XP

1. What advantage does embedding search have over keyword search?

2. Why does the chapter recommend pairing embeddings with keyword search in production?

3. Two documents have nearly identical embeddings, but only one is relevant. Why?

4. Which statement about vector databases is actually true?

5. Why do production RAG systems usually pair vector search with keyword search instead of using vectors alone?

6. Retrieval returns the right chunk, but answers are still wrong. Which is the most likely next thing to check?

7. What is the main reason to preserve document headings when you chunk?

8. Your documents change every day. What is the least disruptive way to keep the index current?

9. Why track cost and p95 latency in the same eval as quality?

10. What should happen when an eval regresses after a change?

11. Why validate that citations point at real chunks?

12. Why attach metadata such as source, section and updated date to every chunk?

What this checkpoint covers

Next section: Tools, Function Calling & Agents