Added integration with Amazon API Gateway + bug fix in _answer_relevance.py (#327)

Closed other PR.
1. Merge with latest changes
2. Added integration with Amazon API Gateway
3. fixed bug in _answer_relevance.py, caused by text empty or NaN in the
cosine_sim calculation

Author: brunopistone

src/ragas/llms/langchain.py

@@ -4,7 +4,7 @@
 
 from langchain.chat_models import AzureChatOpenAI, BedrockChat, ChatOpenAI, ChatVertexAI
 from langchain.chat_models.base import BaseChatModel
-from langchain.llms import AzureOpenAI, Bedrock, OpenAI, VertexAI
+from langchain.llms import AmazonAPIGateway, AzureOpenAI, Bedrock, OpenAI, VertexAI
 from langchain.llms.base import BaseLLM
 from langchain.schema import LLMResult
 
@@ -25,6 +25,9 @@ def isOpenAI(llm: BaseLLM | BaseChatModel) -> bool:
 def isBedrock(llm: BaseLLM | BaseChatModel) -> bool:
     return isinstance(llm, Bedrock) or isinstance(llm, BedrockChat)
 
+def isAmazonAPIGateway(llm: BaseLLM | BaseChatModel) -> bool:
+    return isinstance(llm, AmazonAPIGateway)
+
 
 # have to specify it twice for runtime and static checks
 MULTIPLE_COMPLETION_SUPPORTED = [
@@ -195,6 +198,8 @@ def generate(
         temperature = 0.2 if n > 1 else 1e-8
         if isBedrock(self.llm) and ("model_kwargs" in self.llm.__dict__):
             self.llm.model_kwargs = {"temperature": temperature}
+        elif isAmazonAPIGateway(self.llm) and ("model_kwargs" in self.llm.__dict__):
+            self.llm.model_kwargs = {"temperature": temperature}
         else:
             self.llm.temperature = temperature
 

src/ragas/metrics/_answer_relevance.py

@@ -89,8 +89,11 @@ def _score_batch(
 
             scores = []
             for question, gen_questions in zip(questions, results):
-                cosine_sim = self.calculate_similarity(question, gen_questions)
-                scores.append(cosine_sim.mean())
+                if question is not None and question != "" and len(gen_questions) > 0:
+                    cosine_sim = self.calculate_similarity(question, gen_questions)
+                    scores.append(cosine_sim.mean())
+                else:
+                    scores.append(0.0)
 
         return scores