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

Where do you store embeddings? A vector database. In this episode, we use ChromaDB — a lightweight vector store that embeds documents automatically and lets you search by natural language. We build a recipe knowledge base with 6 recipes, then query it 3 ways: "What can I make with chicken?" finds fried rice. "I want something healthy for breakfast" finds banana smoothie. "How do I make Italian pasta?" finds carbonara with the lowest distance score of 0.547. ChromaDB handles embedding, storage, and search — all in one library. No manual embedding code needed. Student code: https://github.com/GoCelesteAI/build-ai-apps-python/tree/main/episode15 Every keystroke is shown on screen with 3-second pauses so you can follow along at your own pace. What You'll Learn: • ChromaDB — lightweight vector database, runs in memory • Creating a client and collection • Adding documents — ChromaDB embeds them automatically • Querying with natural language — collection.query() • n_results — how many matches to return • Distance scores — lower means more similar • Semantic search vs keyword search • 3 queries proving meaning-based retrieval works • Running Python scripts with :!python % Timestamps: 0:00 - Introduction 0:12 - What is a Vector Store? (Preview) 0:46 - Creating vector_store.py 0:52 - import chromadb 1:12 - Create client and collection 2:02 - Recipe knowledge base (6 recipes) 2:55 - Add recipes to collection — auto-embedded 3:38 - Save progress 3:48 - Query 1: What can I make with chicken? 5:48 - Query 2: I want something healthy for breakfast 7:28 - Query 3: How do I make Italian pasta? 9:32 - Save and run 10:05 - Chicken → Fried Rice! Distance 1.16 10:15 - Healthy breakfast → Banana Smoothie! Semantic match 10:25 - Italian pasta → Carbonara! Lowest distance 0.547 10:40 - Three queries, all correct. Pure meaning, not keywords 10:58 - Code review 11:18 - Recap: 3 Key Takeaways 11:50 - End Screen 1. ChromaDB embeds documents automatically when you add them — no manual embedding code 2. Query with natural language — ChromaDB finds the most similar documents by meaning