fix(tests): accept either initial event type from xAI realtime

xAI's Grok Voice Agent API used to emit 'conversation.created' as the first
event over the WebSocket. It has since shipped a fully OpenAI-compatible
'session.created' event (and may still emit the legacy 'conversation.created'
on some routes), which breaks the strict-equality assertion in the realtime
e2e test:

  AssertionError: Expected conversation.created, got session.created

This is an upstream behavior change, not a regression in our code. Loosen the
base realtime test so 'get_initial_event_type()' may return a tuple of
acceptable event types, and have the xAI subclass accept both
'conversation.created' and 'session.created'. The OpenAI subclasses keep
their single-string contract unchanged.

Co-authored-by: Krrish Dholakia <krrish-berri-2@users.noreply.github.com>

tests/llm_translation/realtime/base_realtime_tests.py

@@ -10,7 +10,7 @@ import json
 import os
 import sys
 from abc import ABC, abstractmethod
-from typing import Optional
+from typing import Optional, Tuple, Union
 
 import pytest
 import websockets
@@ -153,8 +153,14 @@ class BaseRealtimeTest(ABC):
         pass
 
     @abstractmethod
-    def get_initial_event_type(self) -> str:
-        """Return the expected initial event type (e.g., 'session.created' or 'conversation.created')"""
+    def get_initial_event_type(self) -> Union[str, Tuple[str, ...]]:
+        """Return the expected initial event type(s).
+
+        May return a single event type (e.g. ``'session.created'``) or a tuple
+        of acceptable types when the upstream provider can legitimately emit
+        more than one initial event (e.g. xAI's Grok Voice Agent has shipped
+        both ``conversation.created`` and ``session.created``).
+        """
         pass
 
     def get_skip_reason(self) -> str:
@@ -229,9 +235,14 @@ class BaseRealtimeTest(ABC):
 
         # Verify initial event
         initial_event = websocket_client.messages_received[0]
+        expected_event_type = self.get_initial_event_type()
+        if isinstance(expected_event_type, str):
+            allowed_event_types: Tuple[str, ...] = (expected_event_type,)
+        else:
+            allowed_event_types = tuple(expected_event_type)
         assert (
-            initial_event["type"] == self.get_initial_event_type()
-        ), f"Expected {self.get_initial_event_type()}, got {initial_event.get('type')}"
+            initial_event["type"] in allowed_event_types
+        ), f"Expected one of {allowed_event_types}, got {initial_event.get('type')}"
 
     @pytest.mark.asyncio
     async def test_realtime_with_query_params(self):

tests/llm_translation/realtime/test_xai_realtime.py

@@ -7,6 +7,7 @@ Uses the base test class to ensure consistent behavior across providers.
 
 import os
 import sys
+from typing import Tuple
 
 import pytest
 
@@ -19,10 +20,12 @@ class TestXAIRealtime(BaseRealtimeTest):
     """
     E2E tests for xAI Realtime API.
 
-    xAI's Grok Voice Agent API is OpenAI-compatible but uses:
-    - Different initial event: "conversation.created" instead of "session.created"
-    - Different endpoint: wss://api.x.ai/v1/realtime
+    xAI's Grok Voice Agent API is OpenAI-compatible:
+    - Endpoint: wss://api.x.ai/v1/realtime
     - Model: grok-4-1-fast-non-reasoning
+    - Initial event: historically "conversation.created"; xAI has since shipped
+      "session.created" (matching OpenAI). Accept either to avoid spurious
+      failures whenever xAI flips the wire format.
     """
 
     def get_model(self) -> str:
@@ -31,5 +34,5 @@ class TestXAIRealtime(BaseRealtimeTest):
     def get_api_key_env_var(self) -> str:
         return "XAI_API_KEY"
 
-    def get_initial_event_type(self) -> str:
-        return "conversation.created"
+    def get_initial_event_type(self) -> Tuple[str, ...]:
+        return ("conversation.created", "session.created")