# Zero-Shot Text Classification

Text classification with your own labels and no training: topic, intent, urgency, sentiment or anything else, in many languages.

Send a text of up to 4,000 characters and one to ten questions, each a list of labels you make up: which team should handle this, is it urgent, what is the sentiment, which product is it about. You get the chosen label with a score from 0 to 1 for each question, or every label that fits when you ask for more than one. No training and no examples needed, and it reads English, German, French, Spanish, Romanian and many other languages. Answers usually arrive in under a tenth of a second. The model runs on our own hardware in the EU; text is processed in memory and never stored. Scores are a model's estimate: test your labels on your own texts before you automate on them.

- Your own labels, made up on the spot; up to ten questions about one text in a single call.
- Reads English, German, French, Spanish, Romanian and many other languages.
- Usually under a tenth of a second, on our own hardware in the EU; text is never stored.

## Calls

### `POST https://readymadeapi.com/v1/text/classify`

Ask up to ten questions about one text, each with your own labels, and get the chosen label and a score for each.

- Price: $0.0005 per text
- MCP tool: `text_classify_classify`
- Answers in the same call.

Request:

```json
{
  "questions": {
    "team": ["billing", "shipping", "account access"],
    "urgent": ["yes", "no"]
  },
  "text": "I was charged twice for the same order."
}
```

Answer:

```json
{
  "answers": {
    "team": {
      "label": "billing",
      "score": 0.9679
    },
    "urgent": {
      "label": "yes",
      "score": 0.9997
    }
  }
}
```

## Limits

- Up to 4,000 characters and up to 10 questions per call.

## Questions

**What is it for?**

Routing support tickets, intent detection, sentiment analysis, tagging feedback, spotting urgent messages, or any yes/no question about a text.

**How reliable are the scores?**

They are a model's estimate. Test your labels on your own texts before you automate on them; clear, distinct labels work best.

**Which languages?**

English, German, French, Spanish, Romanian and many others; accuracy varies, so test on your own texts.

**How long can the text be?**

Up to 4,000 characters.

## Using it

- Send `Authorization: Bearer <key>` and a JSON body. Keys and credit: https://readymadeapi.com/account
- Every answer carries `X-Request-Id`, `X-Units` and `X-Charged-Usd`.
- An error is a JSON problem with a `title` to act on and a `detail` in plain words; nothing is charged for it.
- Full schema: https://readymadeapi.com/openapi.json
