Why Unitary?

⏺️Detoxify endpoints

Detoxify is an easy-to-use Python library that detects hateful or offensive language.

Detoxify is an open-source tool to help researchers and practitioners identify potential toxic comments.

Detoxify models are trained to predict toxic comments on 3 Jigsaw challenges: Toxic comment classification, Unintended Bias in Toxic comments, and Multilingual toxic comment classification.

Detoxify endpoints

Detoxify is offered as a set of two standalone text-only API endpoints so that you can avoid the hassle of hosting open-source models.

Any requests you send should meet the requirements listed on Features and requirements.

Please follow the API Authentication instructions first. This will generate a token that is valid for 24 hours and must be used in subsequent API requests.

POST endpoint reference

Queue text for Detoxify classification

post

This endpoint predicts the Detoxify categories of text.

Authorizations
Body
callback_urlstring · uri · min: 1 · max: 2083Optional
textstring · max: 262144Required
modelall ofOptionalDefault: multilingual
string · enumOptional

An enumeration.

Possible values:
correlation_idstring · max: 128Optional
Responses
200
Successful Response
application/json
post
POST /detoxify/text/ HTTP/1.1
Host: 
Authorization: Bearer JWT
Content-Type: application/x-www-form-urlencoded
Accept: */*
Content-Length: 99

"callback_url='https://example.com'&text='text'&model='multilingual'&correlation_id='text'"
{
  "job_id": "text"
}

GET endpoint reference

Retrieve the results of a queued Detoxify classification

get

This endpoint returns the results of a previously queued classify job

Authorizations
Path parameters
job_idstring · uuid4Required
Responses
200
Successful Response
application/json
get
GET /detoxify/text/{job_id} HTTP/1.1
Host: 
Authorization: Bearer JWT
Accept: */*

{
  "status": "done",
  "results": {
    "toxicity": 1,
    "severe_toxicity": 1,
    "obscene": 1,
    "threat": 1,
    "insult": 1,
    "identity_attack": 1,
    "sexual_explicit": 1
  },
  "msg": "text"
}

PreviousList of Item and CharacteristicsNextHow to select thresholds

Last updated

Was this helpful?