omni-moderation-latest

Quick example

curl https://reapi.ai/api/v1/moderations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "omni-moderation-latest",
    "input": [
      { "type": "text", "text": "Is this comment safe to publish?" },
      { "type": "image_url", "image_url": { "url": "https://example.com/photo.jpg" } }
    ]
  }'

import requests

resp = requests.post(
    "https://reapi.ai/api/v1/moderations",
    headers={
        "Authorization": "Bearer YOUR_API_KEY",
        "Content-Type": "application/json",
    },
    json={
        "model": "omni-moderation-latest",
        "input": [
            {"type": "text", "text": "Is this comment safe to publish?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}},
        ],
    },
    timeout=30,
)
print(resp.json())

const r = await fetch("https://reapi.ai/api/v1/moderations", {
  method: "POST",
  headers: {
    Authorization: "Bearer YOUR_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "omni-moderation-latest",
    input: [
      { type: "text", text: "Is this comment safe to publish?" },
      { type: "image_url", image_url: { url: "https://example.com/photo.jpg" } },
    ],
  }),
});
console.log(await r.json());

package main

import (
    "bytes"
    "encoding/json"
    "fmt"
    "io"
    "net/http"
)

func main() {
    body, _ := json.Marshal(map[string]any{
        "model": "omni-moderation-latest",
        "input": []map[string]any{
            {"type": "text", "text": "Is this comment safe to publish?"},
            {"type": "image_url", "image_url": map[string]string{"url": "https://example.com/photo.jpg"}},
        },
    })
    req, _ := http.NewRequest("POST",
        "https://reapi.ai/api/v1/moderations", bytes.NewReader(body))
    req.Header.Set("Authorization", "Bearer YOUR_API_KEY")
    req.Header.Set("Content-Type", "application/json")

    resp, _ := http.DefaultClient.Do(req)
    defer resp.Body.Close()
    out, _ := io.ReadAll(resp.Body)
    fmt.Println(string(out))
}

Submit response

{
  "id": "task_01a04de4f56c77f387b030319362ff46",
  "model": "omni-moderation-latest",
  "status": "processing",
  "created_at": 1788013250
}

Poll GET /api/v1/tasks/{id} until status === "completed". A check usually finishes within a few seconds.

Authentication

Every call needs a Bearer token. Generate keys at reapi.ai/settings/apikeys.

Authorization: Bearer YOUR_API_KEY

Endpoint

POST /api/v1/moderations
GET  /api/v1/tasks/{id}

Submission is async. The POST returns immediately with a task id; polling the task endpoint returns the same envelope until completion. Polling does not consume credits.

Request body

model — required

Always "omni-moderation-latest".

input — required

The content to moderate. Three shapes are accepted:

Content blocks:

Strings and text blocks must be non-empty; arrays must have at least one item.

Limits: up to 500 items per request and 30,000 words (300,000 characters) of text in total, summed across every string / text block. Split larger jobs across requests.

stream — optional (boolean)

Only false is supported. Omit it (the default) or send false; any other value is rejected with 20003. The task contract already delivers the whole verdict in one poll response, so there is nothing to stream.

Output

{
  "id": "task_01a04de4f56c77f387b030319362ff46",
  "model": "omni-moderation-latest",
  "status": "completed",
  "created_at": 1788013250,
  "output": {
    "moderation": {
      "results": [
        {
          "flagged": true,
          "categories": {
            "harassment": true,
            "harassment/threatening": false,
            "hate": false,
            "hate/threatening": false,
            "illicit": false,
            "illicit/violent": false,
            "self-harm": false,
            "self-harm/instructions": false,
            "self-harm/intent": false,
            "sexual": false,
            "sexual/minors": false,
            "violence": false,
            "violence/graphic": false
          },
          "category_scores": {
            "harassment": 0.8869,
            "harassment/threatening": 0.0384,
            "hate": 0.0001,
            "hate/threatening": 0.0000,
            "illicit": 0.0000,
            "illicit/violent": 0.0000,
            "self-harm": 0.0005,
            "self-harm/instructions": 0.0000,
            "self-harm/intent": 0.0001,
            "sexual": 0.0000,
            "sexual/minors": 0.0000,
            "violence": 0.0322,
            "violence/graphic": 0.0000
          },
          "category_applied_input_types": {
            "harassment": ["text"],
            "harassment/threatening": ["text"],
            "hate": ["text"],
            "hate/threatening": ["text"],
            "illicit": ["text"],
            "illicit/violent": ["text"],
            "self-harm": ["text", "image"],
            "self-harm/instructions": ["text", "image"],
            "self-harm/intent": ["text", "image"],
            "sexual": ["text", "image"],
            "sexual/minors": ["text"],
            "violence": ["text", "image"],
            "violence/graphic": ["text", "image"]
          }
        }
      ]
    }
  },
  "usage": { "credits": 1 },
  "error": null
}

Reading a result

output.moderation.results[] holds one entry per input item (one per string for an array of strings; a single entry for a content-block array). Each entry:

The 13 categories: harassment, harassment/threatening, hate, hate/threatening, illicit, illicit/violent, self-harm, self-harm/instructions, self-harm/intent, sexual, sexual/minors, violence, violence/graphic.

The category vocabulary and scores are returned exactly as the model produces them; reAPI does not remap or round them.

Pricing

Billed per moderation unit:

• Text: one unit per 1,000 words, summed across every text item in the request. Each text item counts as at least 50 words, so a batch of very short strings is not billed as free.
• Images: one unit per image_url block.
• Every request bills at least one unit.

Units are converted to credits once per request and rounded up, so a large batch is cheaper per item than many single calls. Polling does not add any charge. Failed requests are refunded automatically.

1 credit = $0.001 USD. See the pricing table on the model page for the current per-unit rate.

Errors

Standard envelope:

{
  "error": {
    "code": 20003,
    "message": "input: input is required",
    "request_id": "req_..."
  }
}

Common cases:

Full catalog: /docs/api/errors.

Tips

• Send an array of strings to screen a batch of comments in one task — you get one verdict per string and a single bill.
• Pair text and an image in one content-block array when they are published together (a caption plus its photo); the model scores the combination as one item.
• Treat category_scores as the tunable signal: flagged uses the model's default thresholds, which may be stricter or looser than your policy.
• Image URLs must be fetchable from the public internet. Signed URLs work as long as they are valid when the task runs.

Related

• AI Text Detector — score text for AI authorship.
• AI Humanizer — rewrite AI text to read human.