feat(api): Support input guardrail checks when role: 'system'

Author: christinaexyou

nemoguardrails/server/api.py

@@ -1026,6 +1026,9 @@ async def _process_message(
     if role == "user":
         return None, _create_check_options(run_input=True)
 
+    if role == "system":
+        return None, _create_check_options(run_input=True)
+
     if role == "assistant":
         if "tool_calls" in msg:
             # Tool output rails - validate tool calls before execution
@@ -1039,7 +1042,7 @@ async def _process_message(
         return None, _create_check_options(run_tool_input=True)
 
     # Unsupported role
-    raise ValueError(f"Unsupported message role: '{role}'. Supported roles are: 'user', 'assistant', 'tool'.")
+    raise ValueError(f"Unsupported message role: '{role}'. Supported roles are: 'user', 'system', 'assistant', 'tool'.")
 
 
 def _build_check_messages(role: str, content: str, msg: dict) -> List[dict]:
@@ -1059,6 +1062,9 @@ def _build_check_messages(role: str, content: str, msg: dict) -> List[dict]:
     if role == "user":
         return [{"role": "user", "content": content}]
 
+    if role == "system":
+        return [{"role": "user", "content": content}]
+
     if role == "assistant":
         return [
             {"role": "user", "content": ""},
@@ -1071,7 +1077,7 @@ def _build_check_messages(role: str, content: str, msg: dict) -> List[dict]:
         return [tool_msg]
 
     # This should never be reached since _process_message validates the role first
-    raise ValueError(f"Unsupported message role: '{role}'. Supported roles are: 'user', 'assistant', 'tool'.")
+    raise ValueError(f"Unsupported message role: '{role}'. Supported roles are: 'user', 'system', 'assistant', 'tool'.")
 
 
 @app.post(
@@ -1083,6 +1089,7 @@ async def guardrail_checks(body: GuardrailsChatCompletionRequest, request: Reque
 
     This endpoint validates messages against configured guardrails using role-based routing:
     - user messages: evaluated by input rails
+    - system messages: evaluated by input rails
     - assistant messages: evaluated by output rails
     - tool messages: evaluated by tool_input rails
 
@@ -1111,6 +1118,7 @@ async def process_checks():
 
         Messages are checked independently based on role:
         - user messages: input rails
+        - system messages: input rails
         - assistant messages: output rails
         - tool messages: tool_input rails
         """

tests/test_guardrail_checks_api.py

@@ -338,13 +338,31 @@ def test_no_config_and_no_default():
     api.app.default_config_id = "simple_rails"
 
 
+def test_system_message_passes():
+    """System message triggers input rails and passes."""
+    response = client.post(
+        "/v1/guardrail/checks",
+        json={
+            "model": "test",
+            "messages": [{"role": "system", "content": "You are a helpful assistant"}],
+            "guardrails": {"config_id": "simple_rails"},
+        },
+    )
+
+    result = response.json()
+    assert result["status"] == "success"
+    assert len(result["messages"]) == 1
+    assert result["messages"][0]["role"] == "system"
+    assert result["messages"][0]["index"] == 0
+
+
 def test_unsupported_message_role():
     """Unsupported message role returns error."""
     response = client.post(
         "/v1/guardrail/checks",
         json={
             "model": "test",
-            "messages": [{"role": "system", "content": "You are a helpful assistant"}],
+            "messages": [{"role": "developer", "content": "some content"}],
             "guardrails": {"config_id": "simple_rails"},
         },
     )