Search by meaning needs a different kind of index.
Direct answer
A vector database stores embeddings and quickly finds the nearest ones to a query embedding. It lets AI applications search by meaning rather than keywords and is a common part of retrieval-augmented generation (RAG). For small projects you may not need a dedicated one.
How it works
- Embed your documents and store the vectors.
- Embed the user's question.
- Search for the closest vectors.
- Return the matching documents or pass them to a language model.
When to use one
| Situation | Option |
|---|---|
| Few thousand items | simple in-memory search |
| Existing SQL database | a vector extension |
| Large or fast-growing data | a dedicated vector database |
Things to consider
- Cost and hosting.
- Update speed.
- Filtering by metadata.
- Privacy of the data.
Where Teyro fits
Teyro's The AI Stack course turns this into short lessons you can play in a few minutes a day. It has 100 lessons, and you can start free.
The bottom line
Pick the simplest option that meets your size and speed needs.
