Question 1

What is it?

Accepted Answer

A search store that understands meaning instead of matching words. You hand it your content and it finds the pieces closest in idea to whatever someone asks, even when they phrase it nothing like the original. It is the memory behind search that gets you.

Question 2

What can I build with it?

Accepted Answer

Anything that has to find the right thing by meaning. Smart search over your docs, a chatbot that answers from your own help articles, product and article recommendations, “more like this” suggestions: this is the part that fetches what matters before an answer is written.

Question 3

How does it get set up?

Accepted Answer

You ask your agent and it does the rest: it creates the store, shapes it to fit the content you are working with, and wires it into your app. There are no dashboards to size, no accounts to open, and nothing to keep running.

Question 4

Is my data private?

Accepted Answer

Yes. Your content stays walled off behind the scenes, so nothing you store can be read by anyone else. Even when projects share infrastructure underneath, each one only ever sees its own, with no way to cross the line by accident.

Question 5

How am I billed?

Accepted Answer

Every plan comes with a generous monthly allowance for both adding content and searching it. If you go past it mid-build, it tops up on its own from credit you have added, and that credit never expires, so an indexing run never stalls partway through.

Question 6

What does my agent call?

Accepted Answer

Five HTTP calls over one base URL, with the application key passed as the key query parameter. POST /api/resources/vector-database provisions the store once with a management key. POST /edge/vector-database adds or replaces vectors, each one an id, its values, and whatever metadata you want to filter on later. POST /edge/vector-database/query runs the similarity search: send a vector with topK, includeMetadata, and an optional metadata filter, and matches come back with a score for how well each one fits. POST /edge/vector-database/fetch reads vectors straight back by id, and POST /edge/vector-database/delete removes them by ids, by metadata filter, or all of them.

Question 7

Can I change the profile later?

Accepted Answer

No. The store is provisioned with a profile: dimensions and metric. Dimensions can be 384, 768, 1024, 1536, or 3072, and should match your embedding model; metric is cosine (the default), euclidean, or dotproduct. The profile cannot change afterwards, so a different one means deleting the resource and provisioning again.

Row	plan	vector reads	vector writes
1	Free	1M	500,000
2	Plus	10M	5M
3	Pro	50M	25M

Vector Database

What you get, by default.

Answers from your content

Your content stays yours

Endpoints

Fluid pricing.

Free

Plus

Pro

Questions.