Text & embeddings

Moderations

Check text and images for potentially harmful content.

POST /v1/moderations

Compatible with the OpenAI Moderations API. Classifies input against categories such as harassment, hate, self-harm, sexual content and violence, so you can filter user input or model output.

Authentication

Authorization: Bearer sk-xxxx

Request body

Parameter	Type	Required	Description
`input`	string \| array	Yes	Text, an array of texts, or (for `omni-moderation-*`) an array of `text` / `image_url` parts.
`model`	string	No	`omni-moderation-latest` (text and images) or `text-moderation-latest`. If omitted, a default moderation model is used — specify it explicitly for predictable results.

Request example

cURL

curl https://rutoken.ai/v1/moderations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "omni-moderation-latest",
    "input": "I want to hurt someone."
  }'

Python

from openai import OpenAI

client = OpenAI(base_url="https://rutoken.ai/v1", api_key="sk-xxxx")
resp = client.moderations.create(
    model="omni-moderation-latest",
    input="I want to hurt someone.",
)
print(resp.results[0].flagged)

Node.js

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://rutoken.ai/v1", apiKey: "sk-xxxx" });
const resp = await client.moderations.create({
  model: "omni-moderation-latest",
  input: "I want to hurt someone.",
});
console.log(resp.results[0].flagged);

Image and text together:

json

{
  "model": "omni-moderation-latest",
  "input": [
    {"type": "text", "text": "Check this picture"},
    {"type": "image_url", "image_url": {"url": "https://example.com/image.png"}}
  ]
}

Response

json

{
  "id": "modr-970d409ef3bef3b70c73d8232df86e7d",
  "model": "omni-moderation-latest",
  "results": [
    {
      "flagged": true,
      "categories": {
        "harassment": false,
        "harassment/threatening": false,
        "hate": false,
        "self-harm": false,
        "sexual": false,
        "violence": true,
        "violence/graphic": false
      },
      "category_scores": {
        "harassment": 0.0412,
        "harassment/threatening": 0.0281,
        "hate": 0.0007,
        "self-harm": 0.0002,
        "sexual": 0.0001,
        "violence": 0.8977,
        "violence/graphic": 0.0012
      }
    }
  ]
}

The category list is shortened here.

Field	Type	Description
`results[].flagged`	boolean	`true` if any category is flagged.
`results[].categories`	object	Per-category boolean verdicts.
`results[].category_scores`	object	Per-category confidence scores, `0`–`1`.
`results[].category_applied_input_types`	object	`omni-moderation-*` only: which input types (`text`, `image`) each category was applied to.

← Previous

Rerank

Image generation