diff --git a/litellm/integrations/code_interpreter_interception/handler.py b/litellm/integrations/code_interpreter_interception/handler.py index 950088ff9e0..c83262dbb78 100644 --- a/litellm/integrations/code_interpreter_interception/handler.py +++ b/litellm/integrations/code_interpreter_interception/handler.py @@ -529,6 +529,7 @@ class CodeInterpreterInterceptionLogger(CustomLogger): except Exception: await self._delete_container_for_cache_key(sandbox_key) raise + await self._evict_sandbox_if_a_run_failed(sandbox_key, code_interpreter_calls) optional_params: Final = anthropic_messages_optional_request_params request_patch: Final = AgenticLoopRequestPatch( @@ -583,6 +584,7 @@ class CodeInterpreterInterceptionLogger(CustomLogger): raise tool_messages: Final = [result[0] for result in tool_results] code_interpreter_calls: Final = [result[1] for result in tool_results] + await self._evict_sandbox_if_a_run_failed(sandbox_key, code_interpreter_calls) request_patch: Final = AgenticLoopRequestPatch( model=model, @@ -736,6 +738,12 @@ class CodeInterpreterInterceptionLogger(CustomLogger): except (json.JSONDecodeError, TypeError, AttributeError): return "" + async def _evict_sandbox_if_a_run_failed( + self, sandbox_key: str | None, calls: Sequence[CodeInterpreterCall] + ) -> None: + if any(call["status"] == "failed" for call in calls): + await self._delete_container_for_cache_key(sandbox_key) + async def _sandbox_tool_output(self, sandbox: _Sandbox, arguments: str) -> _ToolRun: match sandbox: case _SandboxUnavailable(): diff --git a/tests/unit/integrations/code_interpreter_interception/test_handler.py b/tests/unit/integrations/code_interpreter_interception/test_handler.py index 1c988a35fdc..59b2fb2f676 100644 --- a/tests/unit/integrations/code_interpreter_interception/test_handler.py +++ b/tests/unit/integrations/code_interpreter_interception/test_handler.py @@ -1511,6 +1511,53 @@ async def test_a_sandbox_failure_answers_the_chat_tool_call_with_a_fixed_failure assert [call["status"] for call in plan.metadata["code_interpreter_calls"]] == ["failed"] +_SESSION_PLAN_SURFACES: Final = ( + pytest.param({}, FakeResponse(output=[_function_call_item()]), id="responses"), + pytest.param( + {_ACTIVE_KEY: True, "_agentic_loop_api_surface": CHAT_COMPLETION_AGENTIC_SURFACE}, + {"choices": [{"message": {"tool_calls": [_chat_function_call_item()]}}]}, + id="chat", + ), +) + + +@pytest.mark.asyncio +@pytest.mark.parametrize(("surface_kwargs", "response"), _SESSION_PLAN_SURFACES) +async def test_a_session_sandbox_whose_code_run_fails_is_not_reused_by_the_next_request(surface_kwargs, response): + sandbox: Final = _SandboxThatCannotRunCode() + logger: Final = CodeInterpreterInterceptionLogger(sandbox_config=sandbox) + + for request_id in ("req1", "req2"): + await logger.async_build_agentic_loop_plan( + tools={ + "tool_calls": [ + { + "id": "call_1", + "call_id": "c1", + "name": LITELLM_CODE_EXECUTION_TOOL_NAME, + "arguments": '{"code":"print(1)"}', + } + ] + }, + model="gpt-5", + messages=[{"role": "user", "content": "x"}], + response=response, + anthropic_messages_provider_config=None, + anthropic_messages_optional_request_params={"tools": []}, + logging_obj=FakeLogging(litellm_call_id=request_id), + stream=False, + kwargs={ + "litellm_call_id": request_id, + _SANDBOX_KEY: "conv-1", + _SESSION_SCOPED_KEY: True, + **surface_kwargs, + }, + ) + + assert len(sandbox.create_calls) == 2 + assert len(sandbox.delete_calls) == 2 + + def _anthropic_message(content: list[dict[str, object]], stop_reason: str) -> dict[str, object]: return { "id": "msg_1",