Question 1

How does a vector DB help a recommendation system?

❌ Optimize SQL queries
❌ Generate new data
✅ Store and retrieve embeddings that represent item features
❌ Store preferences as text

Explanation:
Vector DBs store high-dimensional embeddings that let you find similar items for recommendations.

---

Question 2

Advantage of vector DBs for RAG:

❌ Real-time streaming
❌ Auto-create augmented prompt
✅ Efficient similarity searches using vector embeddings
❌ Visualization tools

Explanation:
RAG relies on vector search to retrieve the most relevant text chunks for the prompt.

---

Question 3

How do similarity searches improve recommendations?

❌ Random recommendations
❌ SQL filtering
❌ Storing tabular data
✅ Identify items close in feature space to user preferences

Explanation:
Similarity search finds items similar to what the user likes, improving accuracy.

---

Question 4

How embedding models + vector DBs work together:

❌ DB creates embeddings
❌ Embedding models visualize data
✅ DB stores embeddings used for similarity search
❌ Embedding models replace DBs

Explanation:
Embedding models create embeddings; vector DBs store and search them.

---

Question 5

What can you not do with modify in Chroma DB?

❌ Change distance metric
❌ Change collection metadata
❌ Add key-value metadata
✅ Change the collection name

Explanation:
Chroma allows metadata/metric updates—but collection name cannot be modified.

---

Question 6

Common pitfall in RAG:

✅ Using a different embedding model for query and document embedding
❌ Using vector DBs
❌ Re-embedding data
❌ Choosing good chunk size

Explanation:
Query + document embeddings must use the same model for similarity search to work correctly.

---

Question 7

Which RAG task is handled outside the vector DB?

❌ Storing embeddings
❌ Retrieving matches
❌ Embedding prompts
✅ Chunking

Explanation:
Chunking happens before storage and is handled by the application layer, not the vector DB.

---

🧾 Summary Table