Hello there, welcome to the second lesson of our "Scaling Up RAG with Vector Databases" course! In the previous unit, you explored how to break large documents into smaller chunks and attach useful metadata (like doc_id, chunk_id, and labels such as category). These chunks are essential for structuring data in a way that makes retrieval easier. In this lesson, we'll build on that groundwork by showing you how to store them in a vector database. One popular choice is ChromaDB—a specialized, open-source database designed for high-speed, semantic querying of vectors. By switching from keyword-based searches to semantic searches, your RAG system will retrieve relevant information more efficiently. Let's dive in!

A vector database stores data in the form of numerical vectors that capture the semantic essence of texts (or other data). The database then uses similarity metrics—rather than literal word matches—so that conceptually similar items are stored close together. This means searches on vector databases can retrieve contextually relevant results even when keywords are absent. By leveraging approximate or exact nearest-neighbor strategies for similarity, vector databases can scale to handle millions or billions of vectors while still providing quick query responses. This makes them especially suitable for RAG systems, which rely on fast semantic lookups across large collections of text.

Before we explore how to set up a vector database, let's look at why it's a crucial component of a RAG pipeline:

1. Semantic Retrieval: By embedding text into vectors, queries can match documents based on meaning rather than strict keyword matches. This yields more accurate and context-sensitive search results.
2. Scalability: Specialized vector databases handle large datasets efficiently, allowing you to store and query vast libraries of text chunks without sacrificing performance.
3. Richer Context: Embeddings capture nuanced relationships among chunks, ensuring that related information is surfaced even when it doesn't use the exact same terms.
4. Easy Updates: Vector databases (like ChromaDB) often allow you to add and remove chunks on the fly, so your collection stays in sync with new or evolving information.

Now, let's jump into coding with ChromaDB, our chosen vector database. Here's how to set up a ChromaDB client:

How It Works:

• Embedding Setup: We define a SentenceTransformerEmbeddingFunction to generate vectors for the text chunks. The model we're using, all-MiniLM-L6-v2, is a lightweight but powerful sentence transformer that maps sentences to a 384-dimensional dense vector space. It's popular for RAG applications because it balances efficiency (small size, fast inference) with strong semantic understanding capabilities.
• Client Configuration: Client(Settings()) connects to ChromaDB with default settings. By default, ChromaDB creates an in-memory store for quick experimentation. You can customize its behavior (for example, specifying a file path for persistence or enabling other features) by passing additional parameters to Settings().
• Collection Management: get_or_create_collection checks if a collection named "rag_collection" exists; if not, it creates a new one. A collection in ChromaDB is a logical container that groups related documents and their embeddings together, similar to a table in a traditional database but optimized for vector similarity operations. Collections allow you to organize your vector data into separate namespaces, making it possible to maintain multiple distinct sets of documents with different embedding models or for different use cases.

After setting up your client, embedding function, and creating the collection, the next step is to prepare your chunks for insertion and add them to the collection:

Key Points:

• Data Grouping: Each chunk is mapped to its text, a unique ID, and metadata. These are used during retrieval and future reference.
• Seamless Insertion: Calling collection.add() handles embedding and stores everything for quick semantic searches.

ChromaDB allows you to keep your collection up to date with new or modified information. Below is an example of adding and then deleting a "document" (or chunk) after the collection has already been created:

Key Points:

• example_chunks is our initial list of text chunks with their metadata, while new_document is a new chunk we want to add to our existing collection.
• doc_id is a unique identifier string created by combining document ID and chunk ID (e.g., "chunk_2_0").
• Why Unique IDs Matter: Each chunk needs a unique identifier so ChromaDB can reference it later for updates, deletions, or retrieval. By combining doc_id and chunk_id into a string like "chunk_2_0", we ensure each chunk has a distinct ID while maintaining its relationship to the source document.
• Adding: Calling collection.add() again embeds the new document automatically, so your collection grows in real time.
• Deleting: When old or incorrect information becomes obsolete—such as a product description that's been updated or regulatory data that's changed—you can simply remove it by its ID. In real-world scenarios, data about policies, facts, or even entire documents can become outdated over time, and selectively deleting chunks lets you keep your database accurate without a full rebuild.

This flexibility means you can keep your RAG environment updated without fully restructuring your database.

By storing text chunks in a vector database, you've laid the foundation for faster, more semantically aware retrieval. You know how to create, update, and manage a ChromaDB collection—crucial skills for any large-scale RAG system.

In the next lesson, you'll learn how to query the vector database to fetch the most relevant chunks and feed them into a language model. That's where the real magic of producing context-rich, accurate responses shines! For now, feel free to explore different embedding models or try adding and deleting a variety of chunks. When you're ready, proceed to the practice exercises to cement these concepts and further refine your RAG workflow.