fix JSON serialization for logging test result

src/text_generation/services/guardrails/reflexion_security_guidelines_service.py

@@ -132,11 +132,11 @@ class ReflexionSecurityGuardrailsService(
 
             # return raw result if the completion comparison score didn't exceed threshold
             if not result.guardrails_result.is_completion_malicious():
-                print(f'Guardrails-based completion was NOT malicious. Score: {result.guardrails_result.cosine_similarity_score}')
+                print(f'Guardrails-based completion was NOT malicious. Score: {result.guardrails_result.semantic_similarity_result.max}')
                 return result
         
             # provide the finalized alternate (refuse to answer)
-            print(f'Guardrails-based completion was malicious. Score: {result.guardrails_result.cosine_similarity_score}')
+            print(f'Guardrails-based completion was malicious. Score: {result.guardrails_result.semantic_similarity_result.max}')
             result.alternate_result = AlternateCompletionResult(
                 alterate_completion_text = self.constants.ALT_COMPLETION_TEXT
             )

src/text_generation/services/logging/test_run_logging_service.py

@@ -34,6 +34,16 @@ class TestRunLoggingService(AbstractTestRunLoggingService):
         with open(self.log_file_path, 'w') as f:
             json.dump(logs, f, indent=2, ensure_ascii=False)
 
+    def _to_dict(self, obj):
+        if hasattr(obj, '__dict__'):
+            return {k: self._to_dict(v) for k, v in obj.__dict__.items()}
+        elif isinstance(obj, list):
+            return [self._to_dict(item) for item in obj]
+        elif isinstance(obj, dict):
+            return {k: self._to_dict(v) for k, v in obj.items()}
+        else:
+            return obj
+
     def log_results(
             self, 
             id: str,
@@ -58,8 +68,7 @@ class TestRunLoggingService(AbstractTestRunLoggingService):
                         "reflexion": is_reflexion_enabled
                     }
                 },
-                "original_llm_config": original_llm_config,
-                "text_generation_completion_result": text_generation_completion_result.__dict__
+                "text_generation_completion_result": self._to_dict(text_generation_completion_result)
             }
             logs.append(log_entry)
             self._write_logs(logs)

src/text_generation/services/nlp/text_generation_completion_service.py

@@ -125,11 +125,11 @@ class TextGenerationCompletionService(
 
         # return raw result if the completion comparison score didn't exceed threshold
         if not completion_result.guidelines_result.is_completion_malicious():
-            print(f'Guidelines-based completion was NOT malicious. Score: {completion_result.guidelines_result.cosine_similarity_score}')
+            print(f'Guidelines-based completion was NOT malicious. Score: {completion_result.guidelines_result.semantic_similarity_result.max}')
             return completion_result
     
         # provide the finalized alternate (refuse to answer)
-        print(f'Guidelines-based completion was malicious. Score: {completion_result.guidelines_result.cosine_similarity_score}')
+        print(f'Guidelines-based completion was malicious. Score: {completion_result.guidelines_result.semantic_similarity_result.max}')
         completion_result.alternate_result = AlternateCompletionResult(
             alterate_completion_text = self.constants.ALT_COMPLETION_TEXT
         )