What the gateway does, shown rather than described

Five task profiles behind one endpoint: text, text extraction from an image, image understanding and speech synthesis. Live demos on the page, with the failure modes named.

Everything below goes through one endpoint. Your code sends a task profile — text, extract, describe, speak — and which model and which provider answer is decided by the gateway, by its own routing rules.

This is not about saving lines of code. It is about a retired model, a raised price or a provider outage no longer being your incident: the route changes inside the gateway and your integration stays as it was. The cost of each request is measured on the way, not guessed afterwards.

POST /api/v1/…

task profiles: 5
endpoints to integrate: one
fallback path: yes
model name in your code: none

From code it is the same endpoint

One field changes — the task profile. Neither the model nor the provider appears in your code, which is why their turnover is not your problem.

Text

{
  "service": "text",
  "text": "Summarise this thread in five points"
}

Image

{
  "service": "ocr",
  "text": "Extract the text as it stands",
  "images": ["<base64>"]
}

Speech

{
  "service": "tts",
  "text": "Your order is packed and waiting"
}

Access is granted one client at a time, after a conversation — there is no open sign-up. That is more honest than a "get your key" form with an unbounded queue behind 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.

Do not enter personal, confidential or client data.
Your text is sent to an external model provider. We do not keep your prompt, but the model's answer stays in our request log for 7 days — we need it to investigate failures. How this works — in the privacy policy.

Or try an example:

Up to 600 characters, no sign-up.

Questions people ask on this page

Are these your own models?

No, and we do not pretend otherwise. What is ours is the gateway: one endpoint, provider and model chosen by task profile, a fallback path when something fails, and per-request cost measured rather than estimated. The models are external and they change without asking us — the point of the gateway is that the change does not reach your code.

Why are the demos closed to search engines?

Because we run working sites for these tasks, and a showcase should not take their queries. This page exists to show what an integration can do, not to solve a visitor's task.

What does one request cost?

It depends on the profile and the provider, and it is measured before a project starts, not after. Text is the cheapest, extraction costs more, image understanding more still, and speech is billed per character.

Can I try it with my own file?

Yes, on the demo pages. Each has its own daily cap: this is a showcase, not a free service, and the cap protects our model bill rather than you. The prepared sample is always available — its answer is cached, so it costs nothing to show.