Skip to main content
OneHot is Sqwish Labs’ API for decisions about text. You send a text, ask questions about it and list the answers you allow. Its models, called kettei, read the text and give every answer a probability. They write no text of their own, so they can’t answer with something you didn’t list.

​
Every answer gets a probability

This support ticket was sent with three questions, one of each decision type. The table shows the answer kettei-core gave, rounded to three places.
Our billing export has failed for two days. We need it for today’s audit.
Each question’s probabilities add up to 1. The likeliest outcome, in bold, comes back as top, and its probability as p_top. You choose what to do with them: act when the likeliest answer is clear, and send the rest to a person. The quickstart sends this request and shows the whole answer.

​
Why decisions, not generated text

  • Only the answers you allow. The model scores your outcomes. There is no free text to parse, and no answer outside your list.
  • A number you can act on. You choose how sure is sure enough, for each question.
  • Several questions, one text. Ask many questions about one text in one request, and pay for the text once.
  • You pay for what the model reads. A decision is billed by its input tokens. The model writes no output, so there are no output tokens to pay for (pricing).
A probability says how the model weighs the answers you listed. It is not a promise that the likeliest one is correct. Test your questions and thresholds on cases whose answers you know before you automate work with them.

​
What you can build

Route a support ticket

Pick the team, spot a deadline and score the impact, in one request.

Analyse every comment

Score sentiment, buying intent and spam on every comment of a post.

Check a tool call before it runs

Run it, ask the user first or refuse, before your agent acts.

Grade retrieved passages

Keep the passages that answer the query, before they reach your model.
The Examples tab has more, each with a request you can run.

​
Try it

The playground runs decisions in the console with no account. Start from a ready-made example and edit the question.

Open the playground

Pick an example, edit the question and see every answer’s probability.
To call the API from your code, sign up in the console and make an API key. New accounts may start on a waitlist (your account).

​
Next steps

Quickstart

Install a client and get your first probabilities.

How OneHot works

The words OneHot uses, and how the pieces fit together.

Decision types

Yes or no, one of several, or a level on a scale.

Coding agents

The plugin and MCP tools for Claude Code, Codex and Cursor.
Your data is not used to train the models offered to anyone else. A model fine-tuned on your data can be called only by your account. For more, see data residency, the privacy policy and the terms of service. Questions, or stuck? Write to help@sqwish.ai.