fix: change default re_flags to NOFLAG in RegexDetector

Author: saichandrapandraju

garak/detectors/function.py

@@ -42,46 +42,11 @@ class FunctionDetector(Detector):
     """Detector that calls user-provided function for custom detection logic.
     
     Inspired by generators.function. The user specifies a function in the format
-    'module#function', and it's called with each output string.
+    'module#function', and it's called with each output string. Additional kwargs 
+    can be passed to the function.
     
-    The function must have signature: function(output_text: str) -> float
+    The function must have signature: function(output_text: str, **kwargs) -> float
     where return value is 0.0 (safe) to 1.0 (detected).
-    
-    Configuration (via --detector_options or --detector_option_file):
-        {
-          "function": {
-            "FunctionDetector": {
-              "function_name": "mymodule#myfunction",
-              "kwargs": {  # Optional: additional args passed to function
-                "threshold": 0.5,
-                "custom_param": "value"
-              }
-            }
-          }
-        }
-    
-    Examples:
-        # In mydetectors.py:
-        def check_length(output_text, min_length=100):
-            return 1.0 if len(output_text) > min_length else 0.0
-        
-        # CLI:
-        --detector_options '{"function": {"FunctionDetector": {
-            "function_name": "mydetectors#check_length",
-            "kwargs": {"min_length": 150}
-        }}}'
-        
-        # In mydetectors.py:
-        def check_pii(output_text):
-            import re
-            has_ssn = bool(re.search(r'\\d{3}-\\d{2}-\\d{4}', output_text))
-            has_email = bool(re.search(r'\\S+@\\S+', output_text))
-            return 1.0 if (has_ssn or has_email) else 0.0
-        
-        # CLI:
-        --detector_options '{"function": {"FunctionDetector": {
-            "function_name": "mydetectors#check_pii"
-        }}}'
     """
 
     DEFAULT_PARAMS = Detector.DEFAULT_PARAMS | {

garak/detectors/regex.py

@@ -1,7 +1,7 @@
 """Regex-based detector for pattern matching in model outputs.
 
 This module provides a configurable regex detector that allows users to
-detect specific patterns in LLM outputs without writing custom detector classes.
+detect specific patterns in LLM outputs.
 
 CLI Examples:
     # Using --config with a consolidated YAML config file (recommended)
@@ -19,7 +19,7 @@
     #           - "api.?key"
     #           - "sk-[A-Za-z0-9]{32,}"
     #         match_type: "any"
-    #         re_flags: ["IGNORECASE"]
+    #         re_flags: ["IGNORECASE"] # list of re.RegexFlag names (NOFLAG by default)
     
     # Or using inline --detector_options
     garak --detectors regex.RegexDetector \\
@@ -42,17 +42,6 @@ class RegexDetector(Detector):
     Useful for detecting specific strings,
     code patterns, credentials, or any text that matches regex patterns.
     
-    Configuration:
-        {
-          "regex": {
-            "RegexDetector": {
-              "patterns": ["pattern1", "pattern2", ...],
-              "match_type": "any",       # or "all"
-              "re_flags": ["IGNORECASE"]  # list of re.RegexFlag names
-            }
-          }
-        }
-    
     Available re_flags (from re.RegexFlag):
         - NOFLAG: No special flags (default behavior)
         - IGNORECASE (or I): Case-insensitive matching
@@ -61,31 +50,13 @@ class RegexDetector(Detector):
         - VERBOSE (or X): Allow comments and whitespace in pattern
         - ASCII (or A): ASCII-only matching for \\w, \\b, etc.
         - LOCALE (or L): Locale-dependent matching (discouraged)
-    
-    Examples:
-        # Detect credentials (case-insensitive)
-        --detector_options '{"regex": {"RegexDetector": {
-            "patterns": ["api.?key", "sk-[A-Za-z0-9]{32,}"]
-        }}}'
-        
-        # Detect multiline patterns (e.g., code blocks)
-        --detector_options '{"regex": {"RegexDetector": {
-            "patterns": ["^import\\s+os", "^sudo"],
-            "re_flags": ["IGNORECASE", "MULTILINE"]
-        }}}'
-        
-        # Case-sensitive exact matching
-        --detector_options '{"regex": {"RegexDetector": {
-            "patterns": ["SECRET_KEY", "API_TOKEN"],
-            "re_flags": ["NOFLAG"]
-        }}}'
     """
     lang_spec = "*"
 
     DEFAULT_PARAMS = Detector.DEFAULT_PARAMS | {
         "patterns": [],  # users must provide patterns
         "match_type": "any",  # "any" or "all"
-        "re_flags": ["IGNORECASE"],  # list of re.RegexFlag names to combine
+        "re_flags": ["NOFLAG"],  # list of re.RegexFlag names to combine
     }
 
     active = False

tests/detectors/test_detectors_regex.py

@@ -321,14 +321,14 @@ def test_invalid_flag_skipped_with_warning(self):
         results = detector.detect(attempt)
         assert results[0] == 1.0  # Case-insensitive match works
 
-    def test_default_flags_ignorecase(self):
-        """Test that default re_flags includes IGNORECASE for backward compatibility"""
+    def test_default_flags_noflag(self):
+        """Test that default re_flags includes NOFLAG"""
         config = {
                 'detectors': {
                     'regex': {
                         'RegexDetector': {
                             'patterns': ['PASSWORD']
-                            # No re_flags specified - should default to IGNORECASE
+                            # No re_flags specified - should default to NOFLAG
                         }
                     }
                 }
@@ -340,7 +340,7 @@ def test_default_flags_ignorecase(self):
         attempt.outputs = [garak.attempt.Message("password")]  # lowercase
 
         results = detector.detect(attempt)
-        assert results[0] == 1.0  # Default is case-insensitive
+        assert results[0] == 0.0  # Default is NOFLAG (case-sensitive)
 
     def test_invalid_match_type_defaults_to_any(self):
         """Test that invalid match_type defaults to 'any' with warning (no error)"""