apply review suggestions

fix

Author: Pouyanpi

nemoguardrails/colang/v2_x/runtime/runtime.py

@@ -20,7 +20,6 @@
 from urllib.parse import urljoin
 
 import aiohttp
-import langchain
 
 from nemoguardrails.actions.actions import ActionResult
 from nemoguardrails.colang import parse_colang_file
@@ -44,8 +43,6 @@
 from nemoguardrails.rails.llm.config import RailsConfig
 from nemoguardrails.utils import new_event_dict, new_readable_uuid
 
-langchain.debug = False
-
 log = logging.getLogger(__name__)
 
 

nemoguardrails/evaluate/evaluate_factcheck.py

@@ -108,7 +108,8 @@ def create_negative_samples(self, dataset):
                 evidence=evidence, answer=answer
             )
             negative_answer = llm_with_config.invoke(formatted_prompt)
-            data["incorrect_answer"] = negative_answer.strip()
+            negative_answer_content = negative_answer.content
+            data["incorrect_answer"] = negative_answer_content.strip()
 
         return dataset
 
@@ -188,14 +189,16 @@ def run(self):
             split="negative"
         )
 
-        print(f"Positive Accuracy: {pos_num_correct/len(self.dataset) * 100}")
-        print(f"Negative Accuracy: {neg_num_correct/len(self.dataset) * 100}")
+        print(f"Positive Accuracy: {pos_num_correct / len(self.dataset) * 100}")
+        print(f"Negative Accuracy: {neg_num_correct / len(self.dataset) * 100}")
         print(
-            f"Overall Accuracy: {(pos_num_correct + neg_num_correct)/(2*len(self.dataset))* 100}"
+            f"Overall Accuracy: {(pos_num_correct + neg_num_correct) / (2 * len(self.dataset)) * 100}"
         )
 
         print("---Time taken per sample:---")
-        print(f"Ask LLM:\t{(pos_time+neg_time)*1000/(2*len(self.dataset)):.1f}ms")
+        print(
+            f"Ask LLM:\t{(pos_time + neg_time) * 1000 / (2 * len(self.dataset)):.1f}ms"
+        )
 
         if self.write_outputs:
             dataset_name = os.path.basename(self.dataset_path).split(".")[0]

nemoguardrails/library/hallucination/actions.py

@@ -71,7 +71,7 @@ async def self_check_hallucination(
                 f"Current LLM engine is {type(llm).__name__}, which may not support all features."
             )
 
-            if "n" not in llm.__fields__:
+            if "n" not in llm.model_fields:
                 log.warning(
                     f"LLM engine {type(llm).__name__} does not support the 'n' parameter for generating multiple completion choices. "
                     f"Please use an OpenAI LLM engine or a model that supports the 'n' parameter for optimal performance."

tests/test_configs/with_custom_llm/custom_llm.py

@@ -20,6 +20,7 @@
     CallbackManagerForLLMRun,
 )
 from langchain_core.language_models.llms import BaseLLM
+from langchain_core.outputs import Generation, LLMResult
 
 
 class CustomLLM(BaseLLM):
@@ -47,9 +48,7 @@ def _generate(
         stop: Optional[List[str]] = None,
         run_manager: Optional[CallbackManagerForLLMRun] = None,
         **kwargs,
-    ):
-        from langchain_core.outputs import Generation, LLMResult
-
+    ) -> LLMResult:
         generations = [
             [Generation(text=self._call(prompt, stop, run_manager, **kwargs))]
             for prompt in prompts
@@ -62,9 +61,7 @@ async def _agenerate(
         stop: Optional[List[str]] = None,
         run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
         **kwargs,
-    ):
-        from langchain_core.outputs import Generation, LLMResult
-
+    ) -> LLMResult:
         generations = [
             [Generation(text=await self._acall(prompt, stop, run_manager, **kwargs))]
             for prompt in prompts