1. What is a Vector Database?

A Vector Database is a database designed to store and search embeddings (vectors) efficiently.

After converting data such as text or images into vectors, we need to store them so we can search and use them later.

For example:

"pink windbreaker" → [0.12, -0.45, 0.78, ...]

A Vector Database stores this vector together with the original data and related information.

2. Why do we need a Vector Database?

Suppose we have 10,000 vectors and want to find the one most similar to a query.

The simplest approach would be:

for vector in database:
    calculate_similarity(query, vector)

This means comparing the query with all 10,000 vectors.

As the number of vectors grows to millions, this becomes slow and expensive.

Vector Databases use indexes and algorithms such as ANN (Approximate Nearest Neighbor) to find similar vectors much faster, without comparing against every vector.

For example:

results = collection.query(
    query_texts=["pink windbreaker"],
    n_results=2
)

The database returns the most relevant results based on vector similarity.

3. What are Vector Databases used for?

Most applications of Vector Databases are related to searching by meaning rather than exact keywords.

Common examples include:

• Recommendation systems: Find similar products and recommend them to users.
• Image search: Find visually similar images.
• Semantic search: Find relevant documents even when they use different words from the query.
• Chatbots / RAG: Store document embeddings so an LLM can retrieve relevant information before generating an answer.

In modern LLM applications, a Vector Database often works as an external knowledge store, allowing the LLM to retrieve information when needed.

4. What are some popular Vector Databases?

Some popular options include:

• ChromaDB: Simple and lightweight, great for learning and POCs.
• Qdrant: Open-source with good performance and suitable for production.
• Weaviate: Supports vector search and hybrid search.
• Milvus: Designed for large-scale vector search.
• Pinecone: Cloud-managed and easy to set up.
• MongoDB Atlas Vector Search: Useful if your application already uses MongoDB.
• PostgreSQL + pgvector: A popular option because you can keep vectors in your existing PostgreSQL database.

You do not always need a dedicated Vector Database. For smaller datasets, you can store vectors in memory or use an existing database such as PostgreSQL with pgvector.