log LLM config

src/text_generation/domain/guidelines_result.py

@@ -1,3 +1,4 @@
+from typing import List
 from src.text_generation.domain.abstract_guidelines_processed_completion import AbstractGuidelinesProcessedCompletion
 from src.text_generation.domain.abstract_text_generation_completion_result import AbstractTextGenerationCompletionResult
 
@@ -6,12 +7,12 @@ class GuidelinesResult(
     AbstractGuidelinesProcessedCompletion):
     def __init__(
             self,
+            completion_text: str,
+            llm_config: dict,
             cosine_similarity_score: float,
-            cosine_similarity_risk_threshold: float,
-            original_completion: str):
+            cosine_similarity_risk_threshold: float):
         
-        self.score = cosine_similarity_score
-        self.original_completion = original_completion
-        self.is_original_completion_malicious = (
-            cosine_similarity_score >= cosine_similarity_risk_threshold
-        )
+        self.completion_text = completion_text
+        self.llm_config = llm_config
+        self.cosine_similarity_score = cosine_similarity_score
+        self.cosine_similarity_risk_threshold = cosine_similarity_risk_threshold

src/text_generation/services/guidelines/base_security_guidelines_service.py

@@ -1,10 +1,11 @@
 from typing import Optional
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.prompts import PromptTemplate, StringPromptTemplate
-from langchain_core.runnables import RunnablePassthrough
+from langchain_core.runnables import RunnablePassthrough, RunnableConfig
 
 from src.text_generation.common.constants import Constants
 from src.text_generation.domain.abstract_guidelines_processed_completion import AbstractGuidelinesProcessedCompletion
+from src.text_generation.domain.guidelines_result import GuidelinesResult
 from src.text_generation.ports.abstract_foundation_model import AbstractFoundationModel
 from src.text_generation.services.guidelines.abstract_security_guidelines_service import AbstractSecurityGuidelinesConfigurationBuilder, AbstractSecurityGuidelinesService
 from src.text_generation.services.nlp.abstract_prompt_template_service import AbstractPromptTemplateService
@@ -51,6 +52,10 @@ class BaseSecurityGuidelinesService(AbstractSecurityGuidelinesService):
         try:
             prompt_template = self._get_template()
             chain = self._create_chain(prompt_template)
-            return chain.invoke(user_prompt)
+            result = GuidelinesResult(
+                completion_text=chain.invoke(user_prompt),
+                llm_config=chain.steps[1].model_dump()
+            )
+            return result
         except Exception as e:
             raise e

src/text_generation/services/logging/test_run_logging_service.py

@@ -41,6 +41,7 @@ class TestRunLoggingService(AbstractTestRunLoggingService):
             is_rag_few_shot_enabled: bool,
             is_cot_enabled: bool,
             is_reflexion_enabled: bool,
+            llm_config: dict,
             scores: List[float],
             mean: float,
             max: float):
@@ -54,6 +55,7 @@ class TestRunLoggingService(AbstractTestRunLoggingService):
                 "is_rag_few_shot_enabled": is_rag_few_shot_enabled,
                 "is_cot_enabled": is_cot_enabled,
                 "is_reflexion_enabled": is_reflexion_enabled,
+                "llm_config": llm_config,
                 "mean": mean,
                 "max": max,
                 "scores": scores

src/text_generation/services/nlp/text_generation_completion_service.py

@@ -1,6 +1,6 @@
 from langchain.prompts import PromptTemplate
 from langchain_core.output_parsers import StrOutputParser
-from langchain_core.runnables import RunnablePassthrough
+from langchain_core.runnables import RunnablePassthrough, RunnableConfig
 
 from src.text_generation.common.constants import Constants
 from src.text_generation.domain.alternate_completion_result import AlternateCompletionResult
@@ -93,17 +93,17 @@ class TextGenerationCompletionService(
             self.example_prompt_injection_completions
         )
         similarity_result: SemanticSimilarityResult = self.semantic_similarity_service.analyze(
-            text=guidelines_result.original_completion
+            text=guidelines_result.completion_text
         )
         
         processed_guidelines_result = GuidelinesResult(
-            original_completion=guidelines_result.original_completion,
+            completion_text=guidelines_result.completion_text,
             cosine_similarity_score=similarity_result.mean,
             cosine_similarity_risk_threshold=self.COSINE_SIMILARITY_RISK_THRESHOLD
         )
         
         completion_result = TextGenerationCompletionResult(
-            original_completion=guidelines_result.original_completion,
+            original_completion=guidelines_result.completion_text,
             guidelines_result=processed_guidelines_result
         )
         
@@ -135,10 +135,11 @@ class TextGenerationCompletionService(
         """Handle: CoT=False, RAG=False"""
         try:
             chain = self._create_chain_without_guidelines()
-            completion = chain.invoke(user_prompt)
-            return TextGenerationCompletionResult(
-                original_completion=completion
+            result = GuidelinesResult(
+                completion_text=chain.invoke(user_prompt),
+                llm_config=chain.steps[1].model_dump()
             )
+            return self._process_guidelines_result(result)
         except Exception as e:
             raise e
 

tests/conftest.py

@@ -33,7 +33,7 @@ from src.text_generation.services.prompt_injection.prompt_injection_example_serv
 from src.text_generation.services.utilities.response_processing_service import ResponseProcessingService
 
 
-MAX_REQUEST_SAMPLE_COUNT = 100
+MAX_REQUEST_SAMPLE_COUNT = 1
 MAX_RESPONSE_SAMPLE_COUNT = 50
 
 def pytest_deselected(items):

tests/integration/test_01_malicious_prompts_no_mitigation.py

@@ -1,6 +1,6 @@
 """
     Usage:
-    $ pytest -k test_00_benign_prompts_no_mitigation -s
+    $ pytest -k test_01_malicious_prompts_no_mitigation -s
 
     Data used:
         1. Samples of compromised output (resulting from NVIDIA garak tests)

tests/integration/test_02_malicious_prompts_cot.py

@@ -1,6 +1,6 @@
 """
     Usage:
-    $ pytest -k test_00_benign_prompts_no_mitigation -s
+    $ pytest -k test_02_malicious_prompts_cot -s
 
     Data used:
         1. Samples of compromised output (resulting from NVIDIA garak tests)

tests/integration/test_03_malicious_prompts_rag.py

@@ -1,6 +1,6 @@
 """
     Usage:
-    $ pytest -k test_00_benign_prompts_no_mitigation -s
+    $ pytest -k test_03_malicious_prompts_rag -s
 
     Data used:
         1. Samples of compromised output (resulting from NVIDIA garak tests)

tests/integration/test_04_malicious_prompts_rag_and_cot.py

@@ -1,6 +1,6 @@
 """
     Usage:
-    $ pytest -k test_00_benign_prompts_no_mitigation -s
+    $ pytest -k test_04_malicious_prompts_rag_and_cot -s
 
     Data used:
         1. Samples of compromised output (resulting from NVIDIA garak tests)

tests/integration/test_utils.py

@@ -50,6 +50,7 @@ def run_prompt_analysis_test(
             is_rag_few_shot_enabled=text_generation_completion_service.is_rag_context_enabled(),
             is_cot_enabled=text_generation_completion_service.is_chain_of_thought_enabled(),
             is_reflexion_enabled=text_generation_completion_service.is_reflexion_enabled(),
+            llm_config=completion_result.guidelines_processed_completion.llm_config,
             scores=result.scores,
             mean=result.mean,
             max=result.max