Include processed_rules to indicate applied inspection rules

aidefense/runtime/inspection_client.py

@@ -118,8 +118,42 @@ def _inspect(self, *args, **kwargs):
         """
         raise NotImplementedError("Subclasses must implement _inspect.")
 
+    @staticmethod
+    def _parse_rule(rule_data: Dict[str, Any]) -> Rule:
+        """
+        Parse a single rule from API response data into a Rule object.
+
+        Args:
+            rule_data (Dict[str, Any]): The rule data from the API response.
+
+        Returns:
+            Rule: The parsed Rule object.
+        """
+        # Try to convert to enum, keep original string if not in enum
+        rule_name = rule_data["rule_name"]
+        try:
+            rule_name = RuleName(rule_data["rule_name"])
+        except ValueError:
+            # Keep the original string for custom rule names
+            pass
+
+        # Try to convert to enum, keep original string if not in enum
+        classification = rule_data.get("classification")
+        try:
+            classification = Classification(rule_data["classification"])
+        except ValueError:
+            # Keep the original string for custom classifications
+            pass
+
+        return Rule(
+            rule_name=rule_name,
+            entity_types=rule_data.get("entity_types"),
+            rule_id=rule_data.get("rule_id"),
+            classification=classification,
+        )
+
     def _parse_inspect_response(
-        self, response_data: Dict[str, Any]
+            self, response_data: Dict[str, Any]
     ) -> "InspectResponse":
         """
         Parse API response (chat or http inspect) into an InspectResponse object.
@@ -149,6 +183,16 @@ def _parse_inspect_response(
                         "classification": "SECURITY_VIOLATION"
                     }
                 ],
+                "processed_rules": [
+                    {
+                        "rule_name": "Prompt Injection",
+                        "rule_id": 0,
+                        "entity_types": [
+                            ""
+                        ],
+                        "classification": "SECURITY_VIOLATION"
+                    }
+                ],
                 "attack_technique": "NONE_ATTACK_TECHNIQUE",
                 "explanation": "",
                 "client_transaction_id": "",
@@ -170,6 +214,14 @@ def _parse_inspect_response(
                         classification=Classification.SECURITY_VIOLATION
                     )
                 ],
+                processed_rules=[
+                    Rule(
+                        rule_name="Prompt Injection",  # Note: This will remain a string since it's not in RuleName enum
+                        rule_id=0,
+                        entity_types=[""],
+                        classification=Classification.SECURITY_VIOLATION
+                    )
+                ],
                 attack_technique="NONE_ATTACK_TECHNIQUE",
                 explanation="",
                 client_transaction_id="",
@@ -191,30 +243,11 @@ def _parse_inspect_response(
                 # Log invalid classification but don't add it
                 self.config.logger.warning(f"Invalid classification type: {cls}")
         # Parse rules if present
-        rules = []
-        for rule_data in response_data.get("rules", []):
-            # Try to convert to enum, keep original string if not in enum
-            rule_name = rule_data["rule_name"]
-            try:
-                rule_name = RuleName(rule_data["rule_name"])
-            except ValueError:
-                # Keep the original string for custom rule names
-                pass
-            # Try to convert to enum, keep original string if not in enum
-            classification = rule_data.get("classification")
-            try:
-                classification = Classification(rule_data["classification"])
-            except ValueError:
-                # Keep the original string for custom classifications
-                pass
-            rules.append(
-                Rule(
-                    rule_name=rule_name,
-                    entity_types=rule_data.get("entity_types"),
-                    rule_id=rule_data.get("rule_id"),
-                    classification=classification,
-                )
-            )
+        rules = [self._parse_rule(rule_data) for rule_data in response_data.get("rules", [])]
+
+        # Parse processed rules if present
+        processed_rules = [self._parse_rule(rule_data) for rule_data in response_data.get("processed_rules", [])]
+
         # Parse severity if present
         severity = None
         try:
@@ -227,6 +260,7 @@ def _parse_inspect_response(
             is_safe=response_data.get("is_safe", True),
             severity=severity,
             rules=rules or None,
+            processed_rules=processed_rules or None,
             attack_technique=response_data.get("attack_technique"),
             explanation=response_data.get("explanation"),
             client_transaction_id=response_data.get("client_transaction_id"),

aidefense/runtime/models.py

@@ -162,6 +162,7 @@ class InspectResponse:
         explanation (Optional[str]): Human-readable explanation of the inspection result.
         client_transaction_id (Optional[str]): Unique client-provided transaction ID for tracing.
         event_id (Optional[str]): Unique event ID assigned by the backend.
+        processed_rules (Optional[List[Rule]]): List of rules applied for inspection.
     """
 
     classifications: List[Classification]
@@ -172,3 +173,4 @@ class InspectResponse:
     explanation: Optional[str] = None
     client_transaction_id: Optional[str] = None
     event_id: Optional[str] = None
+    processed_rules: Optional[List[Rule]] = None