Toxicity

You can get toxicity scores from the /toxicity endpoint or Toxicity class in the Python client. This endpoint/Class takes a single text parameter, which should include the candidate text to be scored for toxicity. The output will include a score that ranges from 0.0 to 1.0. The higher the score, the more toxic the text.

Generate a toxicity score

To generate a toxicity score, you can use the following code examples. Depending on your preference or requirements, select the appropriate method for your application.

1
import os
2
import json
3
import predictionguard as pg
4
5
# Set your Prediction Guard token as an environmental variable.
6
os.environ["PREDICTIONGUARD_TOKEN"] = "<your access token>"
7
8
# Perform the toxicity check.
9
result = pg.Toxicity.check(
10
    		text="This is a perfectly fine statement"
11
)
12
13
print(json.dumps(
14
    result,
15
    sort_keys=True,
16
    indent=4,
17
    separators=(',', ': ')
18
))

The output will look something like:

1
{
2
  "checks": [
3
    {
4
      "index": 0,
5
      "score": 0.00036882987478747964,
6
      "status": "success"
7
    }
8
  ],
9
  "created": 1701721860,
10
  "id": "toxi-REU4ZqFADGAiU6xJmN9PgtlgBO6x9",
11
  "object": "toxicity_check"
12
}