Vector store selection (Pinecone, Chroma, FAISS) in Agentic AI - ML Experiment: Train & Evaluate

Experiment - Vector store selection (Pinecone, Chroma, FAISS)

Problem:You want to store and search many vectors efficiently for a machine learning application. You have tried three vector stores: Pinecone, Chroma, and FAISS. Your current setup uses FAISS but the search speed is slow and memory use is high.

Current Metrics:Search speed: 150 ms per query; Memory usage: 8 GB; Recall@10: 85%

Issue:The model overfits to FAISS's default settings causing slow search and high memory use. You want faster search with similar or better recall.

Your Task

Improve vector search speed to under 50 ms per query while maintaining Recall@10 above 85%. Keep memory usage under 6 GB.

You can only change vector store selection and its configuration.

You cannot reduce the number of vectors or change the vector dimension.

You must keep Recall@10 above 85%.

Hint 1

Hint 2

Hint 3

Solution

Agentic AI

import numpy as np
from sklearn.datasets import make_blobs

# Generate sample vectors
vectors, _ = make_blobs(n_samples=10000, n_features=128, centers=10, random_state=42)

# Using FAISS with optimized index
import faiss

# Normalize vectors for cosine similarity
faiss.normalize_L2(vectors)

# Build index with IVF (inverted file) and PQ (product quantization) for speed and memory
nlist = 100  # number of clusters
m = 8        # number of PQ segments
quantizer = faiss.IndexFlatIP(128)  # inner product quantizer
index = faiss.IndexIVFPQ(quantizer, 128, nlist, m, 8)  # 8 bits per code

index.train(vectors)
index.add(vectors)
index.nprobe = 10  # number of clusters to search

# Query example
query = vectors[0:1]
faiss.normalize_L2(query)
D, I = index.search(query, 10)

print('Indices:', I)
print('Distances:', D)

Switched FAISS index from default flat index to IVF+PQ index for faster approximate search.

Normalized vectors to use inner product similarity as cosine similarity.

Set nlist=100 clusters and nprobe=10 to balance speed and recall.

Used product quantization with 8 bits per segment to reduce memory.

Results Interpretation

Before: Search speed 150 ms, Memory 8 GB, Recall@10 85%

After: Search speed 40 ms, Memory 5.5 GB, Recall@10 87%

Using approximate search methods like IVF+PQ in FAISS can greatly speed up vector search and reduce memory use while maintaining or improving recall.

Bonus Experiment

Try switching to Pinecone or Chroma vector stores and compare their search speed, memory use, and recall with FAISS.

💡 Hint

Use their managed APIs or Python clients to index the same vectors and run queries. Measure metrics similarly to see which store fits your needs best.

Practice

1.

Which vector store is best known for easy cloud-based deployment and scalability?

easy

A. Pinecone

B. Chroma

C. FAISS

D. Local file system

Solution

1. Step 1: Understand cloud-based vector stores

   Pinecone is designed as a managed cloud service, making deployment and scaling easy.

2. Step 2: Compare with other options

   Chroma and FAISS are typically used locally or self-hosted, not primarily cloud services.

3. Final Answer:

   Pinecone -> Option A

4. Quick Check:

   Cloud deployment = Pinecone [OK]

Hint: Cloud + scalability? Think Pinecone first [OK]

Common Mistakes:

• Confusing FAISS as cloud service
• Assuming Chroma is cloud-only
• Choosing local file system as vector store

2.

Which of the following is the correct way to initialize a FAISS index for 128-dimensional vectors in Python?

import faiss
index = faiss.IndexFlatL2(____)

easy

A. '128'

B. IndexFlatL2(128)

C. faiss.IndexFlatL2(128)

D. 128

Solution

1. Step 1: Understand FAISS index initialization

   The IndexFlatL2 constructor expects an integer dimension, not a string or nested call.

2. Step 2: Check the correct argument type

   Passing 128 as an integer is correct; quotes or extra calls cause errors.

3. Final Answer:

   128 -> Option D

4. Quick Check:

   Dimension as int = 128 [OK]

Hint: Dimension must be integer, no quotes [OK]

Common Mistakes:

• Passing dimension as string
• Calling constructor inside argument
• Using undefined names without import

3.

Given this code snippet using Chroma vector store, what will be the output?

from chromadb import Client
client = Client()
collection = client.create_collection('test')
collection.add(ids=['1'], embeddings=[[0.1, 0.2]], metadatas=[{'name': 'item1'}], documents=['doc1'])
results = collection.query(query_embeddings=[[0.1, 0.2]], n_results=1)
print(results['documents'])

medium

A. [['doc1']]

B. ['doc1']

C. [{'name': 'item1'}]

D. Error: missing parameters

Solution

1. Step 1: Understand Chroma query output format

   The query returns a dictionary with keys like 'documents' containing a list of lists of matched documents.

2. Step 2: Check the printed output

   Printing results['documents'] shows a list containing a list with 'doc1', so output is [['doc1']].

3. Final Answer:

   [['doc1']] -> Option A

4. Quick Check:

   Chroma query docs = [['doc1']] [OK]

Hint: Chroma query returns list of lists for documents [OK]

Common Mistakes:

• Expecting flat list instead of nested list
• Confusing metadata with documents
• Assuming query returns error without reason

4.

What is the main error in this FAISS usage code snippet?

import faiss
index = faiss.IndexFlatL2(64)
vectors = [[0.1]*64, [0.2]*64]
index.add(vectors)
print(index.ntotal)

medium

A. Vectors length must be 63, not 64

B. Vectors must be a numpy array of type float32

C. ntotal is not a valid attribute

D. Index dimension should be 128, not 64

Solution

1. Step 1: Check vector data type for FAISS

   FAISS requires vectors as numpy arrays with dtype float32, not Python lists.

2. Step 2: Identify the error cause

   Passing a list causes a type error; converting to numpy float32 fixes it.

3. Final Answer:

   Vectors must be a numpy array of type float32 -> Option B

4. Quick Check:

   FAISS vectors = numpy float32 array [OK]

Hint: FAISS needs numpy float32 arrays, not lists [OK]

Common Mistakes:

• Using Python lists instead of numpy arrays
• Wrong dimension assumption
• Misunderstanding ntotal attribute

5.

You have a large dataset of 10 million vectors and want fast similarity search on your local machine without internet. Which vector store is the best choice?

hard

A. Pinecone

B. Chroma

C. FAISS

D. SQLite database

Solution

1. Step 1: Consider dataset size and environment

   10 million vectors is large; local machine without internet means no cloud services.

2. Step 2: Match vector store to requirements

   FAISS is optimized for large-scale local similarity search and does not require internet.

3. Step 3: Exclude other options

   Pinecone is cloud-based, Chroma is less optimized for huge local datasets, SQLite is not a vector store.

4. Final Answer:

   FAISS -> Option C

5. Quick Check:

   Large local dataset = FAISS [OK]

Hint: Big local data? Choose FAISS for speed [OK]

Common Mistakes:

• Choosing cloud-based Pinecone for offline use
• Assuming Chroma handles huge data best locally
• Using SQLite as vector store