Question 1

Primary purpose of FAISS vs Chroma DB:

✅ FAISS is optimized for efficient similarity search; Chroma DB is a vector database with metadata + LLM integration
❌ Both used the same way
❌ FAISS manages structured data
❌ Both manage structured data

Explanation:
FAISS = similarity search engine
Chroma = full vector DB with metadata, persistence, and LLM tooling.

---

Question 2

When extend FAISS with Milvus instead of using Chroma?

❌ For small datasets
❌ For low-dimensional data
✅ For high-dimensional data needing scalable indexing
❌ For real-time data

Explanation:
Milvus provides distributed, scalable storage on top of FAISS.

---

Question 3

Role of greedy routing in HNSW:

❌ Guarantees exact nearest neighbor
❌ Randomly selects neighbors
❌ Only used during construction
✅ Moves to the neighbor closest to the query to reduce computations

Explanation:
Greedy routing jumps through layers, making ANN search faster.

---

Question 4

Difference in indexing methods:

❌ FAISS = high-dim, Chroma = low-dim
❌ FAISS = HNSW, Chroma = LSH
✅ FAISS supports IVF, Flat, HNSW; Chroma primarily uses HNSW
❌ Same indexing methods

Explanation:
FAISS is more advanced with many index types; Chroma only exposes HNSW.

---

Question 5

Types of indexes used by FAISS:

❌ Only LSH
❌ Only flat & IVF
✅ Flat, IVF, LSH, HNSW
❌ Only LSH & HNSW

Explanation:
FAISS supports multiple index families.

---

Question 6

When choose FAISS over Chroma?

❌ For structured data
❌ For relational DB management
❌ For structured data low-latency
✅ For efficient similarity search in large-scale vector datasets

Explanation:
FAISS is the industry standard for fast similarity search.

---

Question 7

What makes HNSW efficient?

✅ Hierarchical structure enabling fast traversal from coarse → fine
❌ Brute-force search
❌ Compressing vectors
❌ Store all points in top layer

Explanation:
HNSW reduces search time via multi-layer navigation.

---

🧾 Summary Table