Skip to content
HALOWERK

Denkwerk

toxicity score

Scores English texts for toxicity with unitary/unbiased-toxic-roberta (Detoxify "unbiased", RoBERTa-base), run as an int8 ONNX export through transformers.js/ONNX Runtime. Returns sigmoid scores per label (multi-label); texts are scored one by one and truncated at 512 tokens (flagged). English only: texts are checked with franc-min and skipped with a reason when English ranks more than 10 % below the best language. At most 32 texts and 8,000 model tokens per call.

POST https://denk.halowerk.com/v1/toxicity-score

Inputs

Fields of the published input schema
FieldTypeRequiredDescription
textstringNo
textsarrayNo
thresholdnumberNoLabels at or above this score are listed as flagged.
Show full input schema
{
  "type": "object",
  "additionalProperties": false,
  "properties": {
    "text": {
      "type": "string",
      "minLength": 1,
      "maxLength": 5000
    },
    "texts": {
      "type": "array",
      "minItems": 1,
      "maxItems": 32,
      "items": {
        "type": "string",
        "minLength": 1,
        "maxLength": 5000
      }
    },
    "threshold": {
      "type": "number",
      "minimum": 0.01,
      "maximum": 0.99,
      "default": 0.5,
      "description": "Labels at or above this score are listed as flagged."
    }
  }
}

Open original schema

Published input example

{
  "texts": [
    "Thank you so much for your help, have a great day!",
    "You are a stupid idiot and nobody likes you."
  ]
}

Result & documentation

Machine-readable documentation of this hub

Source of this page: the published HALOWERK catalog. Updated: 2026-11-17 UTC.

More endpoints in this hub

All endpoints in this hub โ†’