CalvinBackup/llmsecops-research
1
0
Fork 0
mirror of https://github.com/lightbroker/llmsecops-research.git synced 2026-07-09 22:38:39 +02:00
Code Issues Packages Projects Releases Wiki Activity

implement singular guidelines calls in main service

Browse Source
This commit is contained in:
Adam Wilson
2025-07-18 12:33:51 -06:00
parent c1b4a130f9
commit fec4d711bf
17 changed files with 206 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/integration/test_violation_rate.py
+4 -4
View File
@@ -26,8 +26,8 @@
import pytest
import random
from src.text_generation.domain.guardrails_processed_completion import GuardrailsProcessedCompletion
from src.text_generation.domain.guidelines_processed_completion import GuidelinesProcessedCompletion, TextGenerationCompletionResult
from src.text_generation.domain.guardrails_result import GuardrailsResult
from src.text_generation.domain.guidelines_result import GuidelinesResult, TextGenerationCompletionResult
from src.text_generation.domain.semantic_similarity_result import SemanticSimilarityResult
from src.text_generation.services.guidelines.abstract_generative_ai_security_guidelines_service import AbstractGenerativeAiSecurityGuidelinesService
from src.text_generation.services.logging.test_run_logging_service import TestRunLoggingService
@@ -122,7 +122,7 @@ def test_4_cosine_similarity_blocks_malicious_prompts(
semantic_similarity_service.use_comparison_texts(prompt_injection_sample_responses)
completion_text = text_generation_completion_service.invoke(user_prompt=prompt)
analyzed_response: GuardrailsProcessedCompletion = guardrail_service.process_generated_text(
analyzed_response: GuardrailsResult = guardrail_service.process_generated_text(
model_generated_text=completion_text
)
@@ -154,7 +154,7 @@ def test_5_cosine_similarity_plus_rag_blocks_malicious_output(
semantic_similarity_service.use_comparison_texts(prompt_injection_sample_responses)
completion_text = rag_completion_service.invoke(user_prompt=prompt)
analyzed_response: GuardrailsProcessedCompletion = guardrail_service.process_generated_text(
analyzed_response: GuardrailsResult = guardrail_service.process_generated_text(
model_generated_text=completion_text
)
tests/unit/test_domain.py
+4 -4
View File
@@ -1,13 +1,13 @@
import pytest
from src.text_generation.domain.guardrails_processed_completion import GuardrailsProcessedCompletion
from src.text_generation.domain.guardrails_result import GuardrailsResult
@pytest.mark.unit
def test_guardrail_analyzed_response():
response = GuardrailsProcessedCompletion(
score=0.72839,
response = GuardrailsResult(
cosine_similarity_score=0.72839,
cosine_similarity_risk_threshold=0.5,
original_completion="compromised response",
final="I can't answer that"
guardrails_processed_completion_text="I can't answer that"
)
assert response.is_original_completion_malicious == True