QueryPort

Heading

How it works

01

Your app sends a task profile

Not a model name — a profile: text, ocr, vision, tts. Models get retired; a name hardcoded in an application turns into an unexplained failure the day that happens.

02

A rule picks the provider

Provider order, limits and parameters live in the rule, not in your code. Changing which model serves a profile does not require a deploy on your side.

03

Failure moves to the fallback path

An error, a timeout or a body that says 200 while carrying an error inside — all three are failures, and all three are treated as such. The last one is the dangerous case: it reads as success everywhere that only checks the status code.

04

Every request is measured

Latency, time to first token, tokens in and out, cost, and whether the answer reached a human. Without the last one you only know that the gateway replied, not that anybody saw it.

Try it here

POST /api/v1/probe

Your text goes through the same gateway our products use. Which model answers is decided by a routing rule, not by this page.

Up to 600 characters. No sign-up, nothing is stored.

Requesting access

Private beta. There is no sign-up form and no self-serve key — access is granted case by case, and we would rather tell you that here than after you filled something in. Write to us with four sentences: what you are building, rough volume, which task types, and any hard requirement about where data may go.

info@queryport.com Use the contact form

No published SLA yet, and we will not sign a number we cannot defend with a year of data. The live status block on the home page is what we can show today.

Before you write to us

What do you need from me to decide?
What you are building, roughly what volume you expect, which task types you need (text, OCR, vision, speech) and any hard requirement about where data may go. Four sentences is enough for a first answer.
How long does access take?
We answer within a few working days. If the fit is obvious the key follows quickly; if it is not, we will say no rather than leave you waiting on a maybe.
What does it cost?
During private beta we do not publish a price list, because a price we would have to revise in a month is worse than none. Provider costs are yours and measurable per request; ours is agreed per case.
Can I test before committing?
Yes — that is what beta access is for. There is no contract before you have seen your own traffic go through it.
What are the limits?
Rate limits are set per key and named when the key is issued, not discovered by hitting them. If your expected volume does not fit, we will say so before issuing anything.
What happens if you shut the gateway down?
You point your client at a provider directly. That is the whole migration, and it is deliberate: an integration you cannot leave is a liability, not a feature.