Evaluate Response | HiddenLayer Documentation

curl -X POST https://api.hiddenlayer.ai/detection/v2/response-evaluations \
     -H "HL-Project-Id: internal-search-chatbot" \
     -H "HL-Runtime-Session-Id: sess_4b8cde94604f4c389406a0b2f806069a" \
     -H "Authorization: Bearer <token>" \
     -H "Content-Type: application/json" \
     -d '{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1704067200,
  "model": "gpt-4",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The capital of France is Paris."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 10,
    "total_tokens": 35
  }
}'

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1704067200,
  "model": "gpt-4",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The capital of France is Paris."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 10,
    "total_tokens": 35
  }
}

[BETA] This endpoint is not GA or Production ready and is subject to changes at any time. Breaking changes may occur.

Analyzes an LLM response payload for security threats after it is received from the model.

Accepts any valid provider response payload and returns:

If detect or redact action: the response payload (potentially modified) in the provider’s response format
If block action: a canned block message in the provider’s response format

Use this endpoint inline in your LLM pipeline to evaluate model outputs before returning them to users.

Supported provider formats:

Authentication

AuthorizationBearer

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

Request

This endpoint expects a map from strings to any.

Response headers

HL-Runtime-ActionstringOptional

The action recommended by the policy engine evaluation outcome. It is set to BLOCK when a detection rule is triggered for a block action.

Response

Successful evaluation. Returns the (potentially modified) provider response payload.

Authentication

Headers

Request

Response headers

Response