Skip to main content

​
POST /analyze-prompt

Scans a raw user prompt for injections, toxicity, banned topics, etc.
MethodAuthTime-out
POSTAPI Key5 s

​
Request JSON

{
  "prompt": "string",                  // REQUIRED
  "competitor_list": ["string"] | null,
  "banned_topics":  ["string"] | null,
  "check_safety":   true | false | null,
  "policy":         "string" | null
}
​
prompt
string
required
the text of the prompt sent by the user
​
competitor_list
array of string
if you want to block mentions of any competitors add their names in the request body under this field
​
banned_topics
array of string
if you want to block any topics in the incoming prompt, add them here under this field
​
policy
string
to enforce custom policies on the prompts add them to this field as a string, this would ensure the incoming prompts are abiding by the set policies and will be flagged if they don’t
​
check_safety
boolean
required
having check_safety as True activates the safety guardrails that check for things like toxicity, bias, slurs, violence and more in addition to the existing checks for prompt injection/jailbreak attempts

​
Success — 200 OK

{
  "is_valid": false,
  "recognized_violations": ["TOXIC_PROMPT"],
  "risk_levels": { "TOXIC_PROMPT": "medium" },
  "language_info": { "detected_language": "en", "confidence_score": 0.99 }
}
​
is_valid
boolean
default:"false"
final verdict of the guardrails to convey if the prompt can be allowed or not
​
recognised_violations
array
array of different violations detected in the prompt
​
risk_levels
json
determined risk levels for all the recognised violations. nullable
​
language_info
json
contains info on the detected language for the prompt

​
Common Errors

CodeReason
401Missing / invalid API Key
422Body failed validation
429Rate limit exceeded