Creating Vector Stores

You generated embeddings in Module 6. Now, where do you put them? You need a Vector Database.

Popular Options

1. ChromaDB: Open source, runs locally or in-memory. Great for testing and Python apps.
2. Pinecone: Managed cloud service. Good for massive scale.
3. pgvector: If you use PostgreSQL, this extension adds vector search.

Code Example (ChromaDB)

import chromadb

# 1. Setup client
client = chromadb.Client()
collection = client.create_collection("my_docs")

# 2. Add Documents (Chroma handles embedding call for you if configured, or pass vectors)
collection.add(
    documents=["Doc 1 text...", "Doc 2 text..."],
    metadatas=[{"source": "wiki"}, {"source": "news"}],
    ids=["id1", "id2"]
)

# 3. Query
results = collection.query(
    query_texts=["Search query"],
    n_results=2
)

Summary

A Vector Store is just a database optimized for "Nearest Neighbor" math. It stores the Embedding Vector + The Original Text.

In the next lesson, we discuss Chunking.