fix(llm): avoid asyncio.run crash in Codex complete inside active eve…

core/framework/llm/litellm.py

@@ -9,13 +9,14 @@
 
 import ast
 import asyncio
+import concurrent.futures
 import hashlib
 import json
 import logging
 import os
 import re
 import time
-from collections.abc import AsyncIterator
+from collections.abc import AsyncIterator, Callable
 from datetime import datetime
 from pathlib import Path
 from typing import Any
@@ -705,8 +706,8 @@ def complete(
         # Codex ChatGPT backend requires streaming — delegate to the unified
         # async streaming path which properly handles tool calls.
         if self._codex_backend:
-            return asyncio.run(
-                self.acomplete(
+            return self._run_coro_from_sync(
+                lambda: self.acomplete(
                     messages=messages,
                     system=system,
                     tools=tools,
@@ -783,6 +784,25 @@ def complete(
             raw_response=response,
         )
 
+    @staticmethod
+    def _run_coro_from_sync(coro_factory: Callable[[], Any]) -> Any:
+        """Run a coroutine from sync code, even when an event loop is active.
+
+        ``asyncio.run()`` raises when called from a thread with an active loop
+        (notebooks, async frameworks). In that case, offload coroutine execution
+        to a worker thread with its own event loop.
+        """
+        try:
+            asyncio.get_running_loop()
+        except RuntimeError:
+            return asyncio.run(coro_factory())
+
+        def _run_in_worker() -> Any:
+            return asyncio.run(coro_factory())
+
+        with concurrent.futures.ThreadPoolExecutor(max_workers=1) as executor:
+            return executor.submit(_run_in_worker).result()
+
     # ------------------------------------------------------------------
     # Async variants — non-blocking on the event loop
     # ------------------------------------------------------------------

core/tests/test_litellm_provider.py

@@ -104,6 +104,37 @@ def test_init_ollama_no_key_needed(self):
 class TestLiteLLMProviderComplete:
     """Test LiteLLMProvider.complete() method."""
 
+    def test_complete_codex_backend_in_sync_context(self):
+        """Codex path should work from normal sync callers."""
+        provider = LiteLLMProvider(
+            model="gpt-4o-mini",
+            api_key="test-key",
+            api_base="https://chatgpt.com/backend-api/codex",
+        )
+        expected = LLMResponse(content="codex ok", model="codex-test")
+
+        with patch.object(provider, "acomplete", new=AsyncMock(return_value=expected)) as mock_ac:
+            result = provider.complete(messages=[{"role": "user", "content": "Hello"}])
+
+        assert result == expected
+        mock_ac.assert_awaited_once()
+
+    @pytest.mark.asyncio
+    async def test_complete_codex_backend_in_running_event_loop(self):
+        """Codex path should not crash when complete() is called from async code."""
+        provider = LiteLLMProvider(
+            model="gpt-4o-mini",
+            api_key="test-key",
+            api_base="https://chatgpt.com/backend-api/codex",
+        )
+        expected = LLMResponse(content="codex async ok", model="codex-test")
+
+        with patch.object(provider, "acomplete", new=AsyncMock(return_value=expected)) as mock_ac:
+            result = provider.complete(messages=[{"role": "user", "content": "Hello"}])
+
+        assert result == expected
+        mock_ac.assert_awaited_once()
+
     @patch("litellm.completion")
     def test_complete_basic(self, mock_completion):
         """Test basic completion call."""