Collections, IDs, and Metadata

A vector database is more than vectors

A vector database stores vectors so they can be searched by similarity. Each collection also keeps stable IDs, original document text, and metadata such as source, customer, version, or time. Stable IDs make an update idempotent, meaning that repeating it has the same result instead of adding another copy.

ChromaDB and Qdrant are two vector database tools. Chroma's PersistentClient(path=...) and Qdrant's QdrantClient(path=...) run inside the Python process and save files to local disk. This embedded mode needs no separate background server, network port, or server memory budget in these labs.

Tip: Build an ID from the source path plus chunk number, or from a content hash. A random universally unique identifier (UUID) changes on every retry and creates duplicates.