Introduction

KPD is an AI model that detects hate speech and profanity in Korean text. You send it a sentence, and it tells you whether the sentence is offensive — returning one of three decisions (allow · review · block) along with a score and a severity level.

It's a great fit anywhere people post in real time — community comments, chat, Discord bots. Unlike a simple banned-word list, KPD reads context, so it produces fewer false positives, and it responds in about 50ms on average.

What it can do

  • POST /detect — judge a single sentence
  • POST /detect/batch — judge many sentences at once
  • Operating points (balanced · high_precision · high_recall) to tune sensitivity

How to get started

  1. Send your first request in a few minutes with the Quickstart.
  2. Learn about the free and paid tiers and API keys in Authentication.
  3. Dig into the request and response in Detecting Text and Response Fields.

KPD only returns a judgment. What you do with it — mask, warn, delete — is entirely up to your service.