Policy
Represents a collection of security rules to be applied during a scan. Structure: Dict[PolicyName, PolicyValue]
- The dictionary keys are members of the
PolicyNameenum. - The dictionary values (
PolicyValue) are configuration objects, each requiring at least anactionkey.
Policy-Specific Configurations
The structure of the configuration object depends on the specific policy being configured.| Policy Name | Configuration Schema |
|---|---|
BANNED_TOPICS | {"action": Action, "topics": List[str]} |
SECRETS_KEYS | {"action": Action, "patterns": List[Tuple[str, str]]} |
COMPETITOR_MENTION | {"action": Action, "competitors": List[str]} |
POLICY_VIOLATION | {"action": Action, "rules": List[str]} |
SYSTEM_PROMPT_LEAK | {"action": Action, "system_prompt": str} |
| Other Policies | {"action": Action} |
ApiResult
The structured result from an Argus API scan.
| Key | Type | Description |
|---|---|---|
verdict | Verdict | Required. The definitive outcome of the scan. |
request_id | str | Required. A unique UUID (v4) for the scan request. |
policies_applied | List[AppliedPolicyInfo] | Required. A list detailing every policy that was evaluated by the backend. |
policies_violated | List[ViolatedPolicyInfo] | Required. A list detailing policies that found a violation. Empty if verdict is "passed". |
AppliedPolicyInfo
Details of a policy that was evaluated by the backend. This model is used in the policies_applied list within ApiResult.
| Key | Type | Description |
|---|---|---|
policy_name | PolicyName | The name of the policy that was evaluated. |
action | Action | The configured action for this policy. |
metadata | Metadata | Custom metadata associated with the policy evaluation. |
ViolatedPolicyInfo
Details of a policy that found a violation. This model is used in the policies_violated list within ApiResult.
| Key | Type | Description |
|---|---|---|
policy_name | PolicyName | The name of the policy that found a violation. |
action_taken | Action | The action (FLAG or BLOCK) that was taken as a result of the violation. |
details | Dict[str, Any] | Specific details of the violation, like scores or labels. See schema below. |
details Object Schema by Policy
The structure of the details object within ViolatedPolicyInfo is unique to the policy that was triggered.
| Policy Name | details Schema |
|---|---|
policy_violation_detection | {'reason': str} |
secrets_keys_detection | List[Dict] with {'text': str, 'score': float} |
pii_detection | List[Dict] with {'text': str, 'score': float} |
toxicity_detection | {'label': str, 'score': float} |
competitor_mention_detection | {'detected_competitors': bool} |
banned_topics_detection | {'detected_topics': List[str]} |
prompt_injection_detection | {'score': float} |
unsafe_prompt_protection | {'safety_score': float} |
unsafe_response_detection | {'safety_score': float} |
system_prompt_leak_detection | {'similarity_score': float} |
GuardrailEvent
A comprehensive event object captured by a guardrail. It contains the full context and result of a single analysis.
| Key | Type | Description |
|---|---|---|
node_name | str | The name of the decorated function or node (e.g., ‘Policy_Lookup’). |
payload | str | The actual string content that was analyzed by the guardrail. |
scan_result | ApiResult | The raw ApiResult object returned from the Argus API scan. |
request_policy | Optional[Policy] | The policy configuration sent by the client. |
node_metadata | Optional[Metadata] | User-provided metadata from the decorator for custom observability. |
session_id | Optional[str] | The session ID active during this check. |