CalvinBackup/llmsecops-research
1
0
Fork 0
mirror of https://github.com/lightbroker/llmsecops-research.git synced 2026-07-06 21:17:51 +02:00
Code Issues Packages Projects Releases Wiki Activity

don't send back alternate after guidelines

Browse Source
This commit is contained in:
Adam Wilson
2025-07-28 13:29:21 -06:00
parent 393b76faf4
commit d39d7f1f1f
2 changed files with 1 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/text_generation/services/guardrails/reflexion_security_guidelines_service.py
+1 -1
View File
@@ -142,7 +142,7 @@ class ReflexionSecurityGuardrailsService(
result.alternate_result = None
return result
# provide the finalized alternate (refuse to answer)
# last resort; provide the finalized alternate (refuse to answer)
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/nlp/text_generation_completion_service.py
-4
View File
@@ -128,11 +128,7 @@ class TextGenerationCompletionService(
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.semantic_similarity_result.max}')
completion_result.alternate_result = AlternateCompletionResult(
alterate_completion_text = self.constants.ALT_COMPLETION_TEXT
)
completion_result.finalize_completion_text()
return completion_result