updates for history, streaming handling and fixes

Author: dtam

guardrails/api_client.py

@@ -104,6 +104,8 @@ def stream_validate(
                     )
                 if line:
                     json_output = json.loads(line)
+                    if(json_output.get("error")):
+                        raise Exception(json_output.get("error").get("message"))
                     yield IValidationOutcome.from_dict(json_output)
 
     def get_history(self, guard_name: str, call_id: str):

guardrails/guard.py

@@ -1285,9 +1285,10 @@ def _stream_server_call(
                 guard_history = self._api_client.get_history(
                     self.name, validation_output.call_id
                 )
-                self.history.extend(
-                    [Call.from_interface(call) for call in guard_history]
-                )
+                # TODO renable this. doesnt work in a multiple server environment
+                # self.history.extend(
+                #     [Call.from_interface(call) for call in guard_history]
+                # )
         else:
             raise ValueError("Guard does not have an api client!")
 

guardrails/llm_providers.py

@@ -498,6 +498,9 @@ def _invoke_llm(
             ),
         )
 
+        kwargs.pop("reask_prompt", None)
+        kwargs.pop("reask_instructions", None)
+
         response = completion(
             model=model,
             *args,
@@ -1088,6 +1091,9 @@ async def invoke_llm(
             ),
         )
 
+        kwargs.pop("reask_prompt", None)
+        kwargs.pop("reask_instructions", None)
+
         response = await acompletion(
             *args,
             **kwargs,