Build AI Apps with Python: Search by Meaning with ChromaDB | Episode 15

Build AI Apps with Python: Search by Meaning with ChromaDB

ChromaDB is a lightweight, in-memory vector database that automatically embeds documents and enables natural language search. In this example, we create a small recipe knowledge base and demonstrate how to query it by meaning, finding relevant recipes based on ingredient or meal preferences without manual embedding steps.

Code

from chromadb import Client

# Initialize ChromaDB client and create a collection
client = Client()
collection = client.create_collection(name="recipes")

# Add recipe documents; ChromaDB embeds them automatically
recipes = [
  {"id": "1", "text": "Fried rice with chicken, eggs, and vegetables."},
  {"id": "2", "text": "Banana smoothie with yogurt and honey, a healthy breakfast."},
  {"id": "3", "text": "Classic Italian carbonara pasta with eggs, cheese, and pancetta."},
  {"id": "4", "text": "Vegetable stir fry with tofu and soy sauce."},
  {"id": "5", "text": "Avocado toast with cherry tomatoes and basil."},
  {"id": "6", "text": "Beef tacos with salsa and guacamole."}
]

for recipe in recipes:
  collection.add(documents=[recipe["text"]], ids=[recipe["id"]])

# Query examples: search by natural language meaning
queries = [
  "What can I make with chicken?",
  "I want something healthy for breakfast",
  "How do I make Italian pasta?"
]

for query in queries:
  results = collection.query(query_texts=[query], n_results=1)
  print(f'Query: "{query}"')
  for doc, dist in zip(results["documents"][0], results["distances"][0]):
    print(f'  Result: {doc} (distance: {dist:.3f})')
  print()

Key Points

• ChromaDB automatically generates embeddings for added documents, removing the need for manual embedding code.
• It supports natural language queries to find documents by semantic meaning, not just keyword matching.
• Collections organize documents and enable efficient similarity search in-memory.
• Distance scores indicate how closely query and document embeddings match, with lower values meaning higher relevance.
• ChromaDB is ideal for building lightweight AI-powered search applications with minimal setup.