VectorScope

What does a vector database actually do with your question?

Type a question. Watch it become 384 numbers, get compared with every passage, and come back as ranked results, using the real values from your query.

step one

Ask the corpus a question

Your question is embedded, compared with the passages from the university documents, and sent to the database. Everything below is built from that one run.

try
step two

What happened to your query

The five stages of the real run light up as the backend reports them. Click a stage to watch what happens inside it, using your query's actual numbers: hover any number to see the arithmetic behind it.

Computed locally
Ask a question above to explore what happens to it.
look closer

The same numbers, with tools to dig in. This follows the scene you're on.

All 384 numbers of the query

The query becomes 384 numbers. Browse them all, compare them with the selected passage (dashed outline), and click any one for details.

Run a query to explore all of its embedding values.

step three

What the database returned

Shown in exactly the order the database returned them, with its own scores. VectorScope's similarity is shown next to each for comparison.

Run a query to see the database's results here.

appendix

The earlier benchmark: pgvector vs WaveflowDB

Loading recorded results…