v1/analyze/action/output

This endpoint takes in three parameters:

• action_output: the output received from the action
• action_id: the action id for this action call to the LLM
• source (optional): the source of this action output
• destination (optional) : the destination of this output that will come from the action executor

We return a simple boolean indicating whether a threat was found, and a dictionary of threats found and their relevant metadata.

Request Example

Python

def send_action_output_to_promptarmor(action_output: str, action_id: str, source=None, destination=None, api_key):
    promptarmor_headers = {
        "PromptArmor-Auth": f"Bearer {api_key}",
        "PromptArmor-Session-ID": str(uuid.uuid4()), #The session ID is unique to each user session(e.g. a workflow or conversation)
        "Content-Type": "application/json"
    }

    url = "https://api.aidr.promptarmor.com/v1/analyze/action/output"
    
    data = {
        "action_output": action_output,
        #You should already be tracking this ID to query for the outputs of the action - this is how we link an action to its output
        "action_id": action_id, 
        "source": source,
      	"destination": destination
    }
    
    response = requests.post(url, headers=promptarmor_headers, json=data, verify=True)
    
    print("Detection:", response.json()["detection"])

Response

{
    "detection": false,
    "info": {
        "Code": {
            "detection": false,
            "metadata": {}
        },
        "HTML": {
            "detection": false,
            "metadata": {}
        },
        "HiddenText": {
            "detection": false,
            "metadata": {}
        },
        "InvisibleUnicode": {
            "detection": false,
            "metadata": null
        },
        "Jailbreak": {
            "detection": false
        },
        "MarkdownImage": {
            "detection": false,
            "metadata": null
        },
        "MarkdownURL": {
            "detection": false,
            "metadata": null
        },
        "Secrets": {
            "detection": false,
            "metadata": {}
        },
        "ThreatIntel": {
            "detection": false
        },
        "Anomaly": {
            "detection": false,
            "metadata": {}
        }
    }
}