Vector databases (Pinecone, ChromaDB, Weaviate) in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Vector Database Indexing

Which of the following best describes how vector databases like Pinecone, ChromaDB, and Weaviate index data for fast similarity search?

A They convert data into fixed-length vectors and use approximate nearest neighbor algorithms for indexing.

B They store raw text data and perform keyword matching during queries.

C They use relational tables with primary keys to index data for exact matches.

D They compress images and store them as binary blobs without indexing.

💡 Hint

Think about how similarity search works with numbers instead of text.

Vector databases convert data into vectors (numbers) and use special algorithms to quickly find similar vectors, enabling fast similarity search.

❓ Model Choice

intermediate

Choosing a Vector Database for Real-Time Search

You want to build a real-time recommendation system that updates frequently and requires low latency. Which vector database is best suited for this use case?

A ChromaDB, because it only supports batch updates and slower queries.

B Weaviate, because it does not support vector search but excels in text search.

C Pinecone, because it supports dynamic updates and low-latency queries.

D None of these, because vector databases cannot handle real-time data.

💡 Hint

Consider which database supports fast updates and quick queries.

Pinecone is designed for real-time applications with dynamic updates and low-latency vector search, making it suitable for recommendation systems.

❓ Metrics

advanced

Evaluating Vector Search Quality

Which metric is most appropriate to evaluate the quality of a vector database's approximate nearest neighbor search results?

A BLEU score, measuring similarity between text sequences.

B Precision@k, measuring how many of the top k results are relevant.

C Accuracy, measuring the percentage of correct classifications.

D Mean Squared Error, measuring the difference between predicted and true values.

💡 Hint

Think about how to measure how many returned results are actually relevant.

Precision@k measures how many of the top k retrieved vectors are relevant, which is suitable for evaluating approximate nearest neighbor search quality.

🔧 Debug

advanced

Debugging Vector Search Result Errors

You notice that your vector database returns very poor search results despite correct vector embeddings. Which of the following is the most likely cause?

A The distance metric used for search does not match the embedding space properties.

B The database is missing some raw text data fields.

C The vectors are stored as integers instead of floats, causing syntax errors.

D The database index is too large, causing it to crash on queries.

💡 Hint

Consider how similarity is measured between vectors.

If the distance metric (like Euclidean or cosine) does not match the embedding space, similarity search results will be poor even if embeddings are correct.

❓ Predict Output

expert

Output of Vector Similarity Query Code

What is the output of the following Python code snippet using a vector database client?

Prompt Engineering / GenAI

import numpy as np

# Sample vectors
vectors = {
    'id1': np.array([1.0, 0.0]),
    'id2': np.array([0.0, 1.0]),
    'id3': np.array([1.0, 1.0])
}

# Query vector
query = np.array([1.0, 0.5])

# Function to compute cosine similarity
def cosine_similarity(a, b):
    return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))

# Find id with highest similarity
best_id = max(vectors, key=lambda k: cosine_similarity(query, vectors[k]))
print(best_id)

A id1 and id3 tie

B id1

C id2

D id3

💡 Hint

Calculate cosine similarity for each vector with the query.

Cosine similarity between query and id3 is highest, so 'id3' is printed.