detect async

Author: aditya1503

src/cleanlab_codex/validator.py

@@ -119,8 +119,13 @@ def __init__(
             raise ValueError(error_msg)
 
     def validate(self, query: str, context: str, response: str, prompt=None, form_prompt=None):
-        return asyncio.run(self.validate_async(query, context, response, prompt, form_prompt))
-    
+        try:
+            return asyncio.run(self.validate_async(query, context, response, prompt, form_prompt))
+        except RuntimeError:
+            # If inside an event loop, use `asyncio.create_task()`
+            loop = asyncio.get_event_loop()
+            return loop.create_task(self.validate_async(query, context, response, prompt, form_prompt))
+
     async def validate_async(
         self,
         query: str,
@@ -142,10 +147,11 @@ async def validate_async(
                 - 'is_bad_response': True if the response is flagged as potentially bad (when True, a lookup in Codex is performed), False otherwise.
                 - Additional keys: Various keys from a [`ThresholdedTrustworthyRAGScore`](/cleanlab_codex/types/validator/#class-thresholdedtrustworthyragscore) dictionary, with raw scores from [TrustworthyRAG](/tlm/api/python/utils.rag/#class-trustworthyrag) for each evaluation metric.  `is_bad` indicating whether the score is below the threshold.
         """
-        expert_task = asyncio.create_task(self.remediate_async(query))
-        scores, is_bad_response = await self.detect(query, context, response, prompt, form_prompt)
+        detect_task = self.detect(query, context, response, prompt, form_prompt)
+        remediate_task = self.remediate_async(query)
+        scores, is_bad_response = await detect_task
         if is_bad_response:
-            expert_answer, maybe_entry = await expert_task
+            expert_answer, maybe_entry = await remediate_task
             if expert_answer == None:
                 self._project._sdk_client.projects.entries.add_question(
                     self._project._id, question=query).model_dump()
@@ -158,7 +164,7 @@ async def validate_async(
             **scores,
         }
 
-    def detect(
+    async def detect(
         self,
         query: str,
         context: str,
@@ -180,7 +186,7 @@ def detect(
                 - bool: True if the response is determined to be bad based on the evaluation scores
                   and configured thresholds, False otherwise.
         """
-        scores = self._tlm_rag.score(
+        scores = await self._tlm_rag.score(
             response=response,
             query=query,
             context=context,