From 0ec619e7d114d0f14e50d46f70ea455c833e7767 Mon Sep 17 00:00:00 2001 From: Yuneng Jiang Date: Fri, 28 Aug 2026 09:28:14 -0700 Subject: [PATCH] test: close mutation-testing gaps in container, skills and openai-like config factories Mutation testing surfaced three factory functions whose tests ran against them but asserted nothing that a mutation could break, so every planted bug survived. - litellm/llms/litellm_proxy/skills/code_execution.py: the OpenAI and Anthropic tool schemas were unpinned (the Anthropic one was not reached by any test at all) and the handler's default fallbacks were unchecked - litellm/containers/endpoint_factory.py: the endpoints.json contract, the generated sync/async function set and the response-type mapping were unpinned - litellm/llms/openai_like/dynamic_config.py: the generated Responses API config class had no coverage of auth header, URL resolution or the store override The openai_like tests clear _responses_config_cache around each test. Without that, the module-level cache hands back a class built before the mutation and the tests pass against mutated code. Verified by re-running mutmut per scope: llms/litellm_proxy 45.2% -> 62.8% (70 mutants newly killed) containers 36.8% -> 84.3% (45 mutants newly killed) llms/openai_like 55.7% -> 66.9% (34 mutants newly killed) --- .../containers/test_endpoint_factory.py | 138 +++++++++++++++++ .../skills/test_code_execution.py | 106 +++++++++++++ .../llms/openai_like/test_dynamic_config.py | 144 ++++++++++++++++++ 3 files changed, 388 insertions(+) create mode 100644 tests/test_litellm/containers/test_endpoint_factory.py create mode 100644 tests/test_litellm/llms/litellm_proxy/skills/test_code_execution.py create mode 100644 tests/test_litellm/llms/openai_like/test_dynamic_config.py diff --git a/tests/test_litellm/containers/test_endpoint_factory.py b/tests/test_litellm/containers/test_endpoint_factory.py new file mode 100644 index 00000000000..46564bb1810 --- /dev/null +++ b/tests/test_litellm/containers/test_endpoint_factory.py @@ -0,0 +1,138 @@ +import pytest + +from litellm.containers import endpoint_factory +from litellm.containers.endpoint_factory import ( + RESPONSE_TYPES, + _load_endpoints_config, + create_sync_endpoint_function, + generate_container_endpoints, + get_all_endpoint_names, + get_async_endpoint_names, +) +from litellm.types.containers.main import ( + ContainerFileListResponse, + ContainerFileObject, + DeleteContainerFileResponse, +) + +_SYNC_NAMES = [ + "list_container_files", + "upload_container_file", + "retrieve_container_file", + "delete_container_file", + "retrieve_container_file_content", +] +_ASYNC_NAMES = ["a" + n for n in _SYNC_NAMES] + + +class TestEndpointsConfig: + def test_config_exposes_every_declared_endpoint(self): + config = _load_endpoints_config() + assert [e["name"] for e in config["endpoints"]] == _SYNC_NAMES + + def test_every_endpoint_declares_the_keys_the_factory_reads(self): + for endpoint in _load_endpoints_config()["endpoints"]: + assert set(endpoint) >= { + "name", + "async_name", + "path", + "method", + "path_params", + "response_type", + } + + def test_async_name_is_the_sync_name_prefixed_with_a(self): + for endpoint in _load_endpoints_config()["endpoints"]: + assert endpoint["async_name"] == "a" + endpoint["name"] + + def test_config_is_reread_rather_than_shared_between_callers(self): + first = _load_endpoints_config() + first["endpoints"].clear() + assert len(_load_endpoints_config()["endpoints"]) == len(_SYNC_NAMES) + + +class TestResponseTypeMapping: + def test_mapping_resolves_every_named_response_type(self): + assert RESPONSE_TYPES == { + "ContainerFileListResponse": ContainerFileListResponse, + "ContainerFileObject": ContainerFileObject, + "DeleteContainerFileResponse": DeleteContainerFileResponse, + } + + @pytest.mark.parametrize( + "endpoint_name,expected", + [ + ("list_container_files", ContainerFileListResponse), + ("upload_container_file", ContainerFileObject), + ("retrieve_container_file", ContainerFileObject), + ("delete_container_file", DeleteContainerFileResponse), + ], + ) + def test_each_endpoint_maps_to_its_declared_response_type(self, endpoint_name, expected): + config = next(e for e in _load_endpoints_config()["endpoints"] if e["name"] == endpoint_name) + assert RESPONSE_TYPES[config["response_type"]] is expected + + def test_raw_response_type_is_deliberately_unmapped(self): + config = next( + e for e in _load_endpoints_config()["endpoints"] if e["name"] == "retrieve_container_file_content" + ) + assert config["response_type"] == "raw" + assert RESPONSE_TYPES.get(config["response_type"]) is None + + +class TestGeneratedEndpoints: + def test_generates_exactly_one_sync_and_one_async_function_per_endpoint(self): + assert set(generate_container_endpoints()) == set(_SYNC_NAMES) | set(_ASYNC_NAMES) + + def test_every_generated_value_is_callable(self): + assert all(callable(f) for f in generate_container_endpoints().values()) + + def test_sync_and_async_entries_are_distinct_objects(self): + endpoints = generate_container_endpoints() + for name in _SYNC_NAMES: + assert endpoints[name] is not endpoints["a" + name] + + def test_each_call_builds_fresh_functions(self): + assert ( + generate_container_endpoints()["list_container_files"] + is not generate_container_endpoints()["list_container_files"] + ) + + def test_module_exports_are_wired_and_not_none(self): + for name in _SYNC_NAMES + _ASYNC_NAMES: + assert getattr(endpoint_factory, name) is not None + + +class TestEndpointNameHelpers: + def test_all_endpoint_names_interleaves_sync_then_async_per_endpoint(self): + expected = [n for name in _SYNC_NAMES for n in (name, "a" + name)] + assert get_all_endpoint_names() == expected + + def test_async_endpoint_names_are_only_the_async_ones(self): + assert get_async_endpoint_names() == _ASYNC_NAMES + + def test_async_names_are_a_strict_subset_of_all_names(self): + assert set(get_async_endpoint_names()) < set(get_all_endpoint_names()) + + +class TestSyncEndpointFactory: + def test_returns_a_callable_for_a_minimal_config(self): + assert callable( + create_sync_endpoint_function({"name": "x", "response_type": "ContainerFileObject", "path_params": []}) + ) + + def test_missing_path_params_defaults_to_empty_rather_than_raising(self): + assert callable(create_sync_endpoint_function({"name": "x", "response_type": "ContainerFileObject"})) + + def test_unknown_response_type_is_tolerated_at_build_time(self): + assert callable( + create_sync_endpoint_function({"name": "x", "response_type": "NotARealType", "path_params": []}) + ) + + def test_missing_name_is_a_build_time_error(self): + with pytest.raises(KeyError): + create_sync_endpoint_function({"response_type": "ContainerFileObject"}) + + def test_missing_response_type_is_a_build_time_error(self): + with pytest.raises(KeyError): + create_sync_endpoint_function({"name": "x"}) diff --git a/tests/test_litellm/llms/litellm_proxy/skills/test_code_execution.py b/tests/test_litellm/llms/litellm_proxy/skills/test_code_execution.py new file mode 100644 index 00000000000..d9c4470759f --- /dev/null +++ b/tests/test_litellm/llms/litellm_proxy/skills/test_code_execution.py @@ -0,0 +1,106 @@ +import pytest + +from litellm.llms.litellm_proxy.skills.code_execution import ( + LITELLM_CODE_EXECUTION_TOOL, + CodeExecutionHandler, + LiteLLMInternalTools, + get_litellm_code_execution_tool, + get_litellm_code_execution_tool_anthropic, +) +from litellm.llms.litellm_proxy.skills.constants import ( + DEFAULT_MAX_ITERATIONS, + DEFAULT_SANDBOX_TIMEOUT, +) + +_DESCRIPTION = ( + "Execute Python code in a sandboxed environment. Use this to run code that " + "generates files, processes data, or performs computations. Generated files " + "will be returned directly." +) + + +class TestInternalToolName: + def test_code_execution_tool_name_is_stable(self): + assert LiteLLMInternalTools.CODE_EXECUTION.value == "litellm_code_execution" + + def test_enum_is_str_subclass_so_it_serializes_as_the_bare_name(self): + assert isinstance(LiteLLMInternalTools.CODE_EXECUTION, str) + + +class TestOpenAIToolSchema: + def test_schema_matches_openai_function_tool_contract_exactly(self): + assert get_litellm_code_execution_tool() == { + "type": "function", + "function": { + "name": "litellm_code_execution", + "description": _DESCRIPTION, + "parameters": { + "type": "object", + "properties": {"code": {"type": "string", "description": "Python code to execute"}}, + "required": ["code"], + }, + }, + } + + def test_returns_a_fresh_dict_each_call_so_callers_cannot_mutate_the_shared_one(self): + first = get_litellm_code_execution_tool() + first["function"]["name"] = "clobbered" + assert get_litellm_code_execution_tool()["function"]["name"] == "litellm_code_execution" + + def test_singleton_matches_the_factory(self): + assert LITELLM_CODE_EXECUTION_TOOL == get_litellm_code_execution_tool() + + +class TestAnthropicToolSchema: + def test_schema_matches_anthropic_messages_tool_contract_exactly(self): + assert get_litellm_code_execution_tool_anthropic() == { + "name": "litellm_code_execution", + "description": _DESCRIPTION, + "input_schema": { + "type": "object", + "properties": {"code": {"type": "string", "description": "Python code to execute"}}, + "required": ["code"], + }, + } + + def test_anthropic_shape_is_flat_and_carries_no_openai_only_keys(self): + tool = get_litellm_code_execution_tool_anthropic() + assert "input_schema" in tool + assert "type" not in tool + assert "function" not in tool + assert "parameters" not in tool + + def test_returns_a_fresh_dict_each_call(self): + get_litellm_code_execution_tool_anthropic()["name"] = "clobbered" + assert get_litellm_code_execution_tool_anthropic()["name"] == "litellm_code_execution" + + def test_both_surfaces_agree_on_name_and_description(self): + openai_tool = get_litellm_code_execution_tool() + anthropic_tool = get_litellm_code_execution_tool_anthropic() + assert anthropic_tool["name"] == openai_tool["function"]["name"] + assert anthropic_tool["description"] == openai_tool["function"]["description"] + assert anthropic_tool["input_schema"] == openai_tool["function"]["parameters"] + + +class TestHandlerDefaults: + def test_defaults_come_from_constants_when_nothing_is_passed(self): + handler = CodeExecutionHandler() + assert handler.max_iterations == DEFAULT_MAX_ITERATIONS + assert handler.sandbox_timeout == DEFAULT_SANDBOX_TIMEOUT + + def test_explicit_values_win_over_the_defaults(self): + handler = CodeExecutionHandler(max_iterations=3, sandbox_timeout=7) + assert handler.max_iterations == 3 + assert handler.sandbox_timeout == 7 + + def test_each_argument_falls_back_independently(self): + assert CodeExecutionHandler(max_iterations=3).sandbox_timeout == DEFAULT_SANDBOX_TIMEOUT + assert CodeExecutionHandler(max_iterations=3).max_iterations == 3 + assert CodeExecutionHandler(sandbox_timeout=7).max_iterations == DEFAULT_MAX_ITERATIONS + assert CodeExecutionHandler(sandbox_timeout=7).sandbox_timeout == 7 + + @pytest.mark.parametrize("falsy", [0, None]) + def test_falsy_values_fall_back_to_the_defaults(self, falsy): + handler = CodeExecutionHandler(max_iterations=falsy, sandbox_timeout=falsy) + assert handler.max_iterations == DEFAULT_MAX_ITERATIONS + assert handler.sandbox_timeout == DEFAULT_SANDBOX_TIMEOUT diff --git a/tests/test_litellm/llms/openai_like/test_dynamic_config.py b/tests/test_litellm/llms/openai_like/test_dynamic_config.py new file mode 100644 index 00000000000..55e1a1679de --- /dev/null +++ b/tests/test_litellm/llms/openai_like/test_dynamic_config.py @@ -0,0 +1,144 @@ +import pytest + +from litellm.llms.openai_like import dynamic_config +from litellm.llms.openai_like.dynamic_config import create_responses_config_class +from litellm.llms.openai_like.json_loader import SimpleProviderConfig +from litellm.types.router import GenericLiteLLMParams + +_BASE = {"base_url": "https://api.example.com/v1", "api_key_env": "EXAMPLE_API_KEY"} + + +def _provider(slug, **overrides): + return SimpleProviderConfig(slug=slug, data={**_BASE, **overrides}) + + +@pytest.fixture(autouse=True) +def _isolate_generated_class_cache(): + dynamic_config._responses_config_cache.clear() + yield + dynamic_config._responses_config_cache.clear() + + +class TestClassCaching: + def test_same_slug_returns_the_identical_class_object(self): + provider = _provider("cache_same_slug") + assert create_responses_config_class(provider) is create_responses_config_class(provider) + + def test_cache_is_keyed_on_slug_not_on_the_provider_instance(self): + first = create_responses_config_class(_provider("cache_by_slug")) + second = create_responses_config_class(_provider("cache_by_slug")) + assert first is second + + def test_different_slugs_get_different_classes(self): + assert create_responses_config_class(_provider("cache_slug_a")) is not ( + create_responses_config_class(_provider("cache_slug_b")) + ) + + def test_returns_a_class_not_an_instance(self): + assert isinstance(create_responses_config_class(_provider("returns_class")), type) + + +class TestCustomLlmProvider: + def test_provider_property_reports_the_slug(self): + config = create_responses_config_class(_provider("provider_prop"))() + assert config.custom_llm_provider == "provider_prop" + + +class TestValidateEnvironment: + def test_explicit_api_key_becomes_a_bearer_header(self): + config = create_responses_config_class(_provider("ve_explicit"))() + headers = config.validate_environment( + headers={}, model="m", litellm_params=GenericLiteLLMParams(api_key="sk-explicit") + ) + assert headers["Authorization"] == "Bearer sk-explicit" + + def test_api_key_falls_back_to_the_configured_env_var(self, monkeypatch): + monkeypatch.setenv("VE_ENV_KEY", "sk-from-env") + config = create_responses_config_class(_provider("ve_env", api_key_env="VE_ENV_KEY"))() + headers = config.validate_environment(headers={}, model="m", litellm_params=None) + assert headers["Authorization"] == "Bearer sk-from-env" + + def test_explicit_key_wins_over_the_env_var(self, monkeypatch): + monkeypatch.setenv("VE_LOSER_KEY", "sk-from-env") + config = create_responses_config_class(_provider("ve_precedence", api_key_env="VE_LOSER_KEY"))() + headers = config.validate_environment( + headers={}, model="m", litellm_params=GenericLiteLLMParams(api_key="sk-wins") + ) + assert headers["Authorization"] == "Bearer sk-wins" + + def test_no_key_anywhere_leaves_the_header_unset(self, monkeypatch): + monkeypatch.delenv("VE_MISSING_KEY", raising=False) + config = create_responses_config_class(_provider("ve_missing", api_key_env="VE_MISSING_KEY"))() + assert config.validate_environment(headers={}, model="m", litellm_params=None) == {} + + def test_existing_headers_are_preserved(self): + config = create_responses_config_class(_provider("ve_preserve"))() + headers = config.validate_environment( + headers={"X-Trace": "abc"}, + model="m", + litellm_params=GenericLiteLLMParams(api_key="sk-1"), + ) + assert headers["X-Trace"] == "abc" + + +class TestGetCompleteUrl: + def test_explicit_api_base_gets_the_responses_suffix(self): + config = create_responses_config_class(_provider("url_explicit"))() + assert config.get_complete_url(api_base="https://host/v1", litellm_params={}) == "https://host/v1/responses" + + def test_trailing_slash_is_stripped_before_appending(self): + config = create_responses_config_class(_provider("url_slash"))() + assert config.get_complete_url(api_base="https://host/v1/", litellm_params={}) == "https://host/v1/responses" + + def test_falls_back_to_the_api_base_env_var(self, monkeypatch): + monkeypatch.setenv("URL_BASE_ENV", "https://from-env/v1") + config = create_responses_config_class(_provider("url_env", api_base_env="URL_BASE_ENV"))() + assert config.get_complete_url(api_base=None, litellm_params={}) == "https://from-env/v1/responses" + + def test_falls_back_to_the_configured_base_url_last(self, monkeypatch): + monkeypatch.delenv("URL_UNSET_ENV", raising=False) + config = create_responses_config_class(_provider("url_base_url", api_base_env="URL_UNSET_ENV"))() + assert config.get_complete_url(api_base=None, litellm_params={}) == "https://api.example.com/v1/responses" + + def test_explicit_api_base_wins_over_the_env_var(self, monkeypatch): + monkeypatch.setenv("URL_LOSER_ENV", "https://from-env/v1") + config = create_responses_config_class(_provider("url_precedence", api_base_env="URL_LOSER_ENV"))() + assert ( + config.get_complete_url(api_base="https://explicit/v1", litellm_params={}) + == "https://explicit/v1/responses" + ) + + def test_no_base_anywhere_raises_naming_the_provider(self): + provider = _provider("url_none") + provider.base_url = None + config = create_responses_config_class(provider)() + with pytest.raises(ValueError, match="url_none"): + config.get_complete_url(api_base=None, litellm_params={}) + + +class TestForceStoreFalse: + def test_force_store_false_overrides_the_caller(self): + config = create_responses_config_class( + _provider("store_forced", special_handling={"force_store_false": True}) + )() + params = {"store": True} + config.transform_responses_api_request( + model="m", + input="hi", + response_api_optional_request_params=params, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + assert params["store"] is False + + def test_without_the_flag_the_callers_store_value_is_left_alone(self): + config = create_responses_config_class(_provider("store_untouched"))() + params = {"store": True} + config.transform_responses_api_request( + model="m", + input="hi", + response_api_optional_request_params=params, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + assert params["store"] is True