Safety Evaluation API

Evaluate text content for safety concerns using Constitutional AI.

POST /api/v1/safety/evaluate

Evaluate a piece of text for safety.

Request

curl -X POST "https://api.creed.space/api/v1/safety/evaluate" \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input_text": "Hello, how can I help you today?",
    "context": {},
    "session_id": ""
  }'

Request Body

Field	Type	Required	Description
input_text	string	Yes	The text to evaluate
context	object	No	Additional context (user info, conversation history)
session_id	string	No	Session ID for conversation tracking

Response

{
  "status": "success",
  "evaluation": {
    "risk_score": 0.05,
    "decision": "permit",
    "is_blocked": false,
    "matched_rules": [],
    "detections": [],
    "decision_time_ms": 12.5
  }
}

Response Fields

Field	Type	Description
risk_score	float	Risk level from 0.0 (safe) to 1.0 (dangerous)
decision	string	Action: permit, forbid, divert, or depends
is_blocked	boolean	Whether the content should be blocked
matched_rules	array	List of safety rules that matched
detections	array	Detailed detection information
decision_time_ms	float	Processing time in milliseconds

Decisions Explained

Decision	Meaning	Recommended Action
permit	Content is safe	Process normally
forbid	Content violates policy	Block and inform user
divert	Requires human review	Queue for moderation
depends	Need more context	Request clarification

Example: Detecting Harmful Content

curl -X POST "https://api.creed.space/api/v1/safety/evaluate" \
  -H "X-API-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input_text": "How do I hack into someone'"'"'s account?"}'

Response:

{
  "status": "success",
  "evaluation": {
    "risk_score": 0.85,
    "decision": "forbid",
    "is_blocked": true,
    "matched_rules": ["illegal_activity", "hacking_assistance"],
    "detections": [
      {
        "type": "policy_violation",
        "category": "illegal_activity",
        "confidence": 0.92
      }
    ],
    "decision_time_ms": 15.2
  }
}

Using Context

Provide context for more accurate evaluation:

{
  "input_text": "I want to kill it",
  "context": {
    "conversation_topic": "video_game",
    "user_age": 25,
    "platform": "gaming_chat"
  }
}

Context helps distinguish between:

• "I want to kill it" (about a video game boss) → permit
• "I want to kill it" (ambiguous threat) → divert

Error Responses

400 Bad Request - Invalid input:

{
  "error": "validation_error",
  "message": "input_text is required",
  "details": {
    "field": "input_text",
    "reason": "missing"
  }
}

429 Rate Limited:

{
  "error": "rate_limit_exceeded",
  "message": "Rate limit exceeded: 100 requests per minute",
  "retry_after": 45
}