Question 1

Primary purpose of FAISS vs Chroma DB:

✅ FAISS = efficient similarity search; Chroma DB = vector database with metadata + LLM integration
❌ Both are just vector DBs
❌ FAISS is for structured data
❌ Both manage structured data

Explanation:
FAISS is a search library, Chroma is a full vector database with persistence + metadata.

---

Question 2

When extend FAISS with Milvus?

❌ Small datasets
❌ Low-dimensional data
✅ When high-dimensional data needs scalable indexing
❌ Real-time processing

Explanation:
Milvus adds distributed scalability on top of FAISS.

---

Question 3

Role of greedy routing in HNSW:

❌ Finds exact nearest neighbor
❌ Random neighbor selection
❌ Only used during construction
✅ Moves step-by-step to the closest neighbor, reducing computations

Explanation:
Greedy routing speeds search by climbing down layers toward best candidates.

---

Question 4

Indexing differences between FAISS & Chroma:

❌ FAISS high-dim, Chroma low-dim
❌ FAISS = HNSW, Chroma = LSH
✅ FAISS supports IVF, Flat, LSH, HNSW — Chroma mainly exposes HNSW
❌ Same indexing methods

Explanation:
FAISS offers many index types; Chroma primarily uses HNSW.

---

Question 5

FAISS index types:

❌ Only LSH
❌ Only Flat + IVF
✅ Flat, IVF, LSH, HNSW
❌ Only LSH + HNSW

Explanation:
FAISS offers a wide variety of index structures.

---

Question 6

When choose FAISS over Chroma?

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

Explanation:
FAISS is the best tool for pure vector similarity search at scale.

---

Question 7

Why is HNSW efficient?

✅ Its hierarchical graph enables fast traversal from coarse → fine layers
❌ Brute-force search
❌ Compress vectors
❌ Store all points in top layer

Explanation:
HNSW reduces distance calculations by navigating through multi-layer graphs.

---

🧾 Summary Table