POST
/
api
/
v1
/
judge
/
safety
curl --request POST \
  --url https://api.collinear.ai/api/v1/judge/safety \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
  "conv_prefix": [
    {
      "role": "<string>",
      "content": "<string>"
    }
  ],
  "response": {
    "role": "<string>",
    "content": "<string>"
  },
  "judge_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "space_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
}'
{
  "judgement": 123,
  "extra": {},
  "category": "<string>"
}

​
Parameters

  • conversation: A list of dictionaries representing the conversation in OpenAI format. Each dictionary should have the keys role and content.

  • response: The model’s response in OpenAI format. This is the text that will be judged for safety.

  • judge_id: The unique identifier for the judge. Learn more about obtaining a judge ID.

  • space_id: The unique identifier for your space. Learn more about obtaining a space ID.

​
Response

The API returns a JSON object with the following fields:

  • judgement: Indicates the safety judgement of the model’s output. Possible values are:

    • 1: Safe/Compliant
    • 0: Unsafe/Refusal

  • extra: Contains the raw output from the judge for further inspection.

  • category (Optional): Depending on the judge, this field may contain the category that was flagged for the response.

Authorizations

​
Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
​
conv_prefix
object[]
required
​
response
object
required
​
judge_id
string
required
​
space_id
string
required

Response

200
application/json
Successful Response
​
judgement
integer
required
​
extra
object | null
​
category
string | null

Was this page helpful?

Suggest editsRaise issue