Vector store selection (Pinecone, Chroma, FAISS) in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Choosing a vector store for large-scale real-time search

You want to build a system that handles millions of vectors and requires fast, real-time similarity search with automatic scaling. Which vector store is the best choice?

A Chroma, because it is an open-source vector database designed for small datasets and local use only.

B Pinecone, because it is a managed cloud service that automatically scales and supports real-time search at large scale.

C FAISS, because it is optimized for CPU and GPU and supports large datasets but requires manual scaling.

D Any of the three will perform equally well for large-scale real-time search without extra setup.

💡 Hint

Think about which option offers managed scaling and cloud support for millions of vectors.

Pinecone is a managed vector database service designed for large-scale, real-time similarity search with automatic scaling. FAISS is powerful but requires manual setup and scaling. Chroma is better suited for smaller, local datasets.

❓ Metrics

intermediate

Evaluating vector store retrieval accuracy

You run a similarity search on a vector store and get a list of retrieved items. Which metric best measures how many relevant items are retrieved among the top results?

A Recall, because it measures the fraction of relevant items retrieved out of all relevant items.

B Precision, because it measures the fraction of retrieved items that are relevant.

C Mean Squared Error, because it measures the average squared difference between predicted and true vectors.

D Accuracy, because it measures the overall correct predictions over all queries.

💡 Hint

Recall focuses on how many relevant items you find, not how many you retrieved.

Recall measures the fraction of relevant items retrieved among all relevant items, which is important for evaluating retrieval completeness in vector search.

❓ Predict Output

advanced

Output of FAISS index search code snippet

What is the output of this Python code using FAISS for a simple vector search?

Agentic AI

import numpy as np
import faiss

# Create 5 vectors of dimension 3
vectors = np.array([[1,0,0],[0,1,0],[0,0,1],[1,1,0],[0,1,1]], dtype='float32')

# Build index
index = faiss.IndexFlatL2(3)
index.add(vectors)

# Query vector
query = np.array([[1,0,0]], dtype='float32')

# Search top 2 nearest neighbors
D, I = index.search(query, 2)
print(I[0].tolist())

A [0, 3]

B [0, 1]

C [1, 4]

D [3, 0]

💡 Hint

Remember FAISS returns nearest neighbors by distance. The query is [1,0,0].

The closest vector to [1,0,0] is itself at index 0, then [1,1,0] at index 3 is the next closest by Euclidean distance.

❓ Model Choice

advanced

Selecting vector store for offline batch processing

You want to perform offline batch similarity searches on a dataset of 100,000 vectors without needing real-time responses. Which vector store is most suitable?

A Pinecone, because it is optimized for real-time cloud queries and automatic scaling.

B None of these, you should use a relational database for vector search.

C FAISS, because it is optimized for fast batch processing on large datasets locally or on GPU.

D Chroma, because it is designed for small datasets and local use only.

💡 Hint

Consider which tool is best for offline batch processing and large datasets without cloud dependency.

FAISS is designed for efficient similarity search on large datasets locally or on GPU, making it ideal for offline batch processing.

🔧 Debug

expert

Identifying error in Pinecone vector insertion code

What error will this Pinecone vector insertion code raise?

Agentic AI

import pinecone
pinecone.init(api_key='fake_key', environment='us-west1-gcp')
index = pinecone.Index('example-index')
vectors = [(1, [0.1, 0.2, 0.3]), (2, [0.4, 0.5])]  # Note second vector has length 2
index.upsert(vectors)

A No error; vectors will be inserted successfully.

B KeyError because the index name 'example-index' does not exist.

C TypeError because vectors must be a dictionary, not a list of tuples.

D ValueError due to inconsistent vector dimensions in the upsert list.

💡 Hint

Check if all vectors have the same dimension length.

Pinecone requires all vectors to have the same dimension length. The second vector has length 2 instead of 3, causing a ValueError.