Post scoring API

Score social media posts

A fast, multilingual classifier that returns 15 attribute scores per post (toxicity, insult, threat, identity attack and eleven others), trained to follow Perspective's scores on Bluesky text. Access is for researchers and developers, and we approve requests by hand.

Request access   Sign in

How it works

Send posts with your key and get scores between 0 and 1 back.

curl -X POST https://models.dignityindifference.org/api/v1/score \
  -H "Authorization: Bearer YOUR_KEY" -H "Content-Type: application/json" \
  -d '{"texts": ["first post", "second post"]}'

Limits are counted in posts per second, so one request with 50 posts costs 50. Your plan sets the rate. When you exceed it you get a 429 with a Retry-After header.

What to know before relying on it

The scores imitate Perspective, so they carry its blind spots and biases. They are not a judgement of harm. About 97% of the training posts were English, so non-English scores are less certain. We do not store the text you send, only counts and timings.