# Named Entity Extraction (NER)

NER with your own labels: name up to 30 kinds of thing in plain words and get every match with its position and score.

Send up to 100,000 characters of text and 1 to 30 labels in plain words: order number, product, city, medication, job title, anything. You get every match with its label, its position in the text and a score from 0 to 1. No training and no examples needed, and it reads many languages. Use it to fill forms from emails, tag documents, or feed a search index. The model runs on our own hardware in the EU; text is processed in memory and never stored. It can miss things and can be wrong: set the threshold to trade one for the other.

- Your labels, not a fixed PERSON / ORG / LOC set: order number, medication, job title, anything.
- No training and no examples. It reads English, German, French, Spanish, Romanian and many other languages.
- Up to 100,000 characters per call, processed in memory on our own hardware in the EU and never stored.

## Calls

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

Send text and the kinds of thing to look for, in your own words, and get each one found with its position and a score.

- Price: $0.001 per 1k characters, up to 100 per call
- MCP tool: `entity_extract_extract`
- Answers in the same call.

Request:

```json
{
  "labels": ["order number", "product", "city", "person", "date"],
  "text": "Order 88213 for two Aeron chairs ships from Rotterdam to Dr. Elena Radu on 3 December."
}
```

Answer:

```json
{
  "entities": [
    {
      "label": "order number",
      "start": 6,
      "end": 11,
      "score": 0.9999,
      "text": "88213"
    },
    {
      "label": "product",
      "start": 20,
      "end": 32,
      "score": 0.9835,
      "text": "Aeron chairs"
    },
    {
      "label": "city",
      "start": 44,
      "end": 53,
      "score": 1.0,
      "text": "Rotterdam"
    },
    {
      "label": "person",
      "start": 57,
      "end": 71,
      "score": 0.9858,
      "text": "Dr. Elena Radu"
    },
    {
      "label": "date",
      "start": 75,
      "end": 85,
      "score": 0.9993,
      "text": "3 December"
    }
  ],
  "counts": {
    "city": 1,
    "date": 1,
    "order number": 1,
    "person": 1,
    "product": 1
  }
}
```

## Limits

- Up to 100,000 characters and 1 to 30 labels per call.

## Questions

**What is it for?**

Filling forms from emails, tagging documents, pulling named things out for a search index, and other NLP jobs that need the things named in a text.

**Which languages?**

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

**How do I get fewer wrong matches?**

Raise the threshold. Every match carries a score from 0 to 1; a higher threshold finds less and is right more often.

**How is a call counted?**

Per 1,000 characters of text, rounded up.

## 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
