AI Vector Index Management: Pinecone, Weaviate, Qdrant, pgvector

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-tools-AI-vector-index-management-creators

1. Which combination of factors does the lesson identify as the primary basis for choosing a vector database?

   1. Whether the database offers a free tier for individual developers
   2. Database popularity ranking on developer forums
   3. The programming language used to build the application
   4. Scale of data, hybrid-search requirements, and existing technology stack

2. For a small-scale application that already uses PostgreSQL as its primary database, which vector database does the lesson recommend as a companion option?

   1. Pinecone
   2. Weaviate
   3. Qdrant
   4. pgvector

3. In the context of vector databases, what does ANN stand for?

   1. Authenticated Vector Notation
   2. Adaptive Node Navigation
   3. Approximate Nearest Neighbor
   4. Advanced Neural Network

4. Why does the lesson characterize re-indexing as an expensive operation?

   1. Because re-indexing millions of vectors takes hours and costs money
   2. Because it requires purchasing new hardware
   3. Because vector databases automatically charge per re-index operation
   4. Because re-indexing requires deleting the entire database first

5. Which two databases are explicitly named in the lesson as serving ANN queries at scale?

   1. Pinecone and Qdrant
   2. Qdrant and pgvector
   3. Weaviate and pgvector
   4. Pinecone and Weaviate

6. What type of search combines vector similarity with traditional attribute or keyword filtering?

   1. Graph traversal search
   2. Semantic search only
   3. Hybrid search
   4. Full-text search

7. In the benchmark metrics mentioned in the lesson, what does recall@10 measure?

   1. The cost of processing 10 queries per second
   2. The percentage of correct results found in the top 10 returned items
   3. The memory usage for storing 10 vectors
   4. The time it takes to return 10 results

8. Why might migrating from one vector database to another become costly, beyond just monetary expenses?

   1. Migration is always free because data can be exported easily
   2. Vector databases do not support data migration
   3. Vector databases require annual licensing fees
   4. Different databases use different embedding formats, requiring reprocessing

9. What is metadata filtering in vector search?

   1. Applying traditional filters (date, category, author) alongside vector similarity
   2. Encrypting vector data for privacy
   3. Converting text metadata into vector representations
   4. Filtering vectors by their file size

10. The lesson states there is no 'universal winner' among vector databases. What does this imply about the database selection process?

    1. Pinecone is always the best choice
    2. All vector databases are essentially the same
    3. The cheapest option should always be chosen
    4. Selection should be based on specific project requirements rather than assuming one is always best

11. What characteristic makes pgvector suitable for certain production scenarios despite not being designed for massive scale?

    1. It requires no setup or configuration
    2. It integrates directly with existing PostgreSQL installations
    3. It is the fastest vector database available
    4. It uses a proprietary cloud-only architecture

12. What trade-off must teams accept when using approximate nearest neighbor (ANN) algorithms instead of exact search?

    1. Lower recall in exchange for faster search speeds
    2. Slower query speeds and higher costs
    3. Loss of metadata support
    4. Inability to filter results

13. The lesson recommends versioning embedding models and indexes together. What problem does this practice prevent?

    1. API rate limiting
    2. Inability to reproduce exact search results after model updates
    3. Database connection timeouts
    4. Running out of storage space

14. What does ingestion throughput measure in vector database benchmarking?

    1. How fast the database returns query results
    2. How quickly vectors can be added to the database
    3. The maximum number of concurrent users
    4. The amount of disk space used per vector

15. In a RAG (Retrieval-Augmented Generation) production workload, when would hybrid search capabilities become particularly important?

    1. When users want both keyword-precise and semantically similar results
    2. When the application only needs semantic similarity
    3. When the database contains fewer than 100 vectors
    4. When using only metadata filters without vectors