> ## Documentation Index
> Fetch the complete documentation index at: https://read.sqwish.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Ask structured questions about your data. Get answers your application can use directly.

D1 is Sqwish’s structured prediction API. Send a text or JSON context with a set
of questions, then use the answers to classify, route, or score that context.

For example, give D1 a customer message and ask whether it needs support, which
team should handle it, and how urgent it is — all in one request.

## Three question types

| Type       | Use it to                             | Answer                                                                   |
| ---------- | ------------------------------------- | ------------------------------------------------------------------------ |
| **Binary** | Evaluate a yes/no question            | `probability_true`, from 0 to 1                                          |
| **Choice** | Select from options you define        | A probability for each option and the `selected` option ID               |
| **Score**  | Rate something against ordered levels | A numeric `value`; the baseline uses the expected zero-based level index |

Mix question types in the same request. Answers preserve your question IDs and
order, so they are straightforward to map back into your application.

## Make your first prediction

Create an API key in the console, send a prediction, and read its `answers`.
The [quickstart](/quickstart) walks through a complete request.

<Note>
  Access is currently by invitation. The staging deployment returns simulated
  answers for integration testing. Check `simulated` in `GET /v1/models` before
  interpreting results; sample scores and latency do not describe a real model.
</Note>

The **API reference** documents each endpoint, including request fields,
responses, and code examples. Read [API behavior](/api) when integrating billing,
retries, or pagination.