scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-11-13 08:52:02 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 174

test: wrap the new developer cache mark test lines at 120 characters

Browse source
This commit is contained in:
mateo-berri 2026-11-08 11:13:22 -08:00
parent bb0ede847a
commit 4ac3beafca
2 changed files with 62 additions and 31 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

77
tests/unit/integrations/test_anthropic_cache_control_hook.py
View file

@ -1319,7 +1319,9 @@ def test_cache_control_hook_never_counts_fewer_marks_than_anthropic_receives_aft
pytest.param("invoke/us.anthropic.claude-opus-5-5", "claude-opus-5-5", id="provider-left-out-of-the-params"),
pytest.param("anthropic/my-claude-alias", "my-claude-alias", id="model-the-cost-map-does-not-know"),
pytest.param(
"anthropic/my-unreadable-claude-alias", "my-unreadable-claude-alias", id="cost-map-entry-the-lookup-cannot-read"
"anthropic/my-unreadable-claude-alias",
"my-unreadable-claude-alias",
id="cost-map-entry-the-lookup-cannot-read",
),
],
)
@ -1406,7 +1408,10 @@ _SECOND_SEARCH_CALL: Final = {**_SEARCH_CALL, "id": "srvtoolu_02Search"}
_CLIENT_MARKED_CONVERSATION: Final = (
{"role": "system", "content": [{"type": "text", "text": "Handbook", "prompt_cache_breakpoint": {"mode": "explicit"}}]},
{
"role": "system",
"content": [{"type": "text", "text": "Handbook", "prompt_cache_breakpoint": {"mode": "explicit"}}],
},
{"role": "user", "content": [{"type": "text", "text": "First", "prompt_cache_breakpoint": {"mode": "explicit"}}]},
{"role": "assistant", "content": "First answer"},
{"role": "user", "content": "Second"},
@ -1495,7 +1500,9 @@ def test_tool_config_injection_point_leaves_client_prompt_cache_breakpoints_on_c
pytest.param("openai/gpt-5.6", {"api_base": "https://gateway.example.com/v1"}, id="reads-them"),
pytest.param("openai/gpt-4o", {}, id="openai-model-that-predates-them"),
pytest.param("hosted_vllm/qwen3", {"api_base": "http://vllm.example.com/v1"}, id="self-hosted"),
pytest.param("azure_ai/qa-unflagged-gpt", {"api_base": "https://qa.services.ai.azure.com/models"}, id="foundry"),
pytest.param(
"azure_ai/qa-unflagged-gpt", {"api_base": "https://qa.services.ai.azure.com/models"}, id="foundry"
),
],
)
def test_cache_control_injection_points_keep_client_prompt_cache_breakpoints_on_a_route_that_does_not_translate_them(
@ -1832,11 +1839,14 @@ def test_cache_control_hook_does_not_overwrite_existing_cache_control():
([{"type": "thinking", "thinking": "hm", "signature": "s"}], []),
([{"type": "text", "text": "Checking."}, "not a content block"], [0]),
],
ids=["text-before-trailing-redacted-thinking", "redacted-thinking-only", "signed-thinking-only", "text-before-a-string"],
ids=[
"text-before-trailing-redacted-thinking",
"redacted-thinking-only",
"signed-thinking-only",
"text-before-a-string",
],
)
def test_assistant_injection_never_marks_a_thinking_part(
content: list[dict[str, str] | str], marked_parts: list[int]
):
def test_assistant_injection_never_marks_a_thinking_part(content: list[dict[str, str] | str], marked_parts: list[int]):
"""Claude 400s a cache_control on a thinking or redacted_thinking block, and Bedrock Converse 400s a cache
point after reasoning, so the breakpoint goes on the last part that can carry one."""
messages: Final[list[AllMessageValues]] = [
@ -5193,7 +5203,6 @@ def test_cache_control_hook_injects_next_to_a_malformed_message_instead_of_raisi
]
def test_system_injection_point_filling_the_cap_without_a_developer_message_logs_no_limit_warning(
monkeypatch: pytest.MonkeyPatch, caplog: pytest.LogCaptureFixture
):
@ -5302,6 +5311,7 @@ def test_developer_prompt_caching_check_answers_without_a_model_lookup(custom_ll
assert list(checked) == [developer, question]
assert injected == [developer, question]
_SHARED_FOUNDRY_MODEL: Final = "azure_ai/qa-shared-gpt-deployment"
_SIBLING_BREAKPOINT_FLAGS: Final = MappingProxyType(
{
@ -5632,7 +5642,9 @@ async def test_a_wildcard_deployments_prompt_cache_flag_leaves_each_matched_mode
await _send_unmarked_system_prompt(router, model, "acompletion")
return b"prompt_cache_options" in chat.calls.last.request.content
assert {model: await sends_prompt_cache_options(model) for model in ("openai/gpt-4.1-qa", "openai/gpt-99.0-qa")} == {
assert {
model: await sends_prompt_cache_options(model) for model in ("openai/gpt-4.1-qa", "openai/gpt-99.0-qa")
} == {
"openai/gpt-4.1-qa": False,
"openai/gpt-99.0-qa": True,
}
@ -5763,7 +5775,9 @@ def test_a_gateway_deployment_declaring_the_flag_picks_openais_dialect_for_a_mod
] == [True, False]
def test_a_gateway_deployment_declaring_the_flag_picks_openais_dialect_when_its_model_info_lists_that_provider() -> None:
def test_a_gateway_deployment_declaring_the_flag_picks_openais_dialect_when_its_model_info_lists_that_provider() -> (
None
):
assert [
AnthropicCacheControlHook.targets_openai_prompt_cache_breakpoint(
"qa-self-hosted-gpt",
@ -5773,10 +5787,9 @@ def test_a_gateway_deployment_declaring_the_flag_picks_openais_dialect_when_its_
for listed_provider in ("hosted_vllm", "openrouter")
] == [True, False]
def test_a_custom_handler_provider_declaring_the_flag_keeps_the_anthropic_mark() -> None:
litellm.register_model(
{"qa-custom-handler/qa-model": {"litellm_provider": "qa-custom-handler", "mode": "chat"}}
)
litellm.register_model({"qa-custom-handler/qa-model": {"litellm_provider": "qa-custom-handler", "mode": "chat"}})
assert (
AnthropicCacheControlHook.targets_openai_prompt_cache_breakpoint(
@ -5978,7 +5991,9 @@ def test_cache_control_hook_reads_a_non_string_answer_id_as_answering_no_tool_ca
)
_DEVELOPER_PREFIX: Final = "Policy handbook rule 8.3: record every backup decision in the tracker within 4 business days."
_DEVELOPER_PREFIX: Final = (
"Policy handbook rule 8.3: record every backup decision in the tracker within 4 business days."
)
_CLAUDE_ON_CONVERSE: Final = "bedrock/converse/us.anthropic.claude-haiku-5-5"
_OFFLINE_AWS_CREDENTIALS: Final[Mapping[str, str]] = MappingProxyType(
{
@ -6174,7 +6189,9 @@ def test_system_injection_point_leaves_a_developer_prompt_unmarked_when_its_mark
{"role": "developer", "content": [{"type": "text", "text": _DEVELOPER_PREFIX}]},
{
"role": "user",
"content": [question_part if question_mark is None else {**question_part, "cache_control": question_mark}],
"content": [
question_part if question_mark is None else {**question_part, "cache_control": question_mark}
],
},
],
tools=[dict(_WEATHER_TOOL) if tool_mark is None else {**_WEATHER_TOOL, "cache_control": tool_mark}],
@ -6207,7 +6224,9 @@ _TOOL_WITH_A_CACHE_CONTROL_ARGUMENT: Final = MappingProxyType(
("developer_mark", "points"),
[
pytest.param(dict(_ONE_HOUR), None, id="client_mark"),
pytest.param(None, [{"location": "message", "role": "system", "control": dict(_ONE_HOUR)}], id="injection_point"),
pytest.param(
None, [{"location": "message", "role": "system", "control": dict(_ONE_HOUR)}], id="injection_point"
),
],
)
def test_one_hour_developer_prompt_stays_cached_beside_a_tool_argument_named_cache_control(
@ -6242,9 +6261,15 @@ _FIVE_MINUTE_TOOL: Final = MappingProxyType({**_WEATHER_TOOL, "cache_control": d
@pytest.mark.parametrize(
("outside_fields", "keeps_the_mark"),
[
pytest.param({"tools": [dict(_TOOL_WITH_A_CACHE_CONTROL_ARGUMENT)]}, True, id="tool_argument_named_cache_control"),
pytest.param(
{"tools": [{**_WEATHER_TOOL, "function": {**_WEATHER_TOOL["function"], "cache_control": dict(_FIVE_MINUTES)}}]},
{"tools": [dict(_TOOL_WITH_A_CACHE_CONTROL_ARGUMENT)]}, True, id="tool_argument_named_cache_control"
),
pytest.param(
{
"tools": [
{**_WEATHER_TOOL, "function": {**_WEATHER_TOOL["function"], "cache_control": dict(_FIVE_MINUTES)}}
]
},
False,
id="mark_under_the_tool_function",
),
@ -6312,7 +6337,9 @@ def _server_search_replaying(results_field: str) -> dict[str, object]:
"messages": [
{
"role": "user",
"content": [{"type": "text", "text": "Which rule covers backups?", "cache_control": dict(_ONE_HOUR)}],
"content": [
{"type": "text", "text": "Which rule covers backups?", "cache_control": dict(_ONE_HOUR)}
],
}
]
}
@ -6346,7 +6373,9 @@ def test_system_injection_point_marks_a_developer_prompt_for_openai_prompt_cache
"object": "chat.completion",
"created": 0,
"model": "gpt-5.6",
"choices": [{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}],
"choices": [
{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}
],
"usage": {"prompt_tokens": 30, "completion_tokens": 4, "total_tokens": 34},
},
)
@ -6550,7 +6579,9 @@ async def test_system_injection_point_leaves_a_developer_prompt_for_a_routed_gem
"object": "chat.completion",
"created": 0,
"model": "gemini-3.8-flash",
"choices": [{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}],
"choices": [
{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}
],
"usage": {"prompt_tokens": 30, "completion_tokens": 4, "total_tokens": 34},
},
)
@ -6586,7 +6617,9 @@ def test_system_injection_point_marks_a_developer_prompt_for_a_gateway_alias_wit
"object": "chat.completion",
"created": 0,
"model": "claude-haiku-5-5",
"choices": [{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}],
"choices": [
{"index": 0, "message": {"role": "assistant", "content": "Rule 8.3."}, "finish_reason": "stop"}
],
"usage": {"prompt_tokens": 30, "completion_tokens": 4, "total_tokens": 34},
},
)

16
tests/unit/llms/base_llm/chat/test_provider_chat_translation.py
View file

@ -791,7 +791,9 @@ def test_developer_message_cache_control_survives_the_system_role_translation(
@pytest.mark.parametrize(
("case", "unmarked_system"),
[
pytest.param(_BY_ID["anthropic_sonnet45"], [{"type": "text", "text": "Be a good bot!"}], id="anthropic_sonnet45"),
pytest.param(
_BY_ID["anthropic_sonnet45"], [{"type": "text", "text": "Be a good bot!"}], id="anthropic_sonnet45"
),
pytest.param(
_BY_ID["bedrock_invoke_haiku"], [{"type": "text", "text": "Be a good bot!"}], id="bedrock_invoke_haiku"
),
@ -963,10 +965,10 @@ _BEDROCK_CONVERSE_SYSTEMS: Final = pytest.param(
)
@pytest.mark.parametrize(("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS])
@pytest.mark.parametrize(
("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS]
("user_marks", "keeps_developer_mark"), [(3, True), (4, False)], ids=["at_limit", "over_limit"]
)
@pytest.mark.parametrize(("user_marks", "keeps_developer_mark"), [(3, True), (4, False)], ids=["at_limit", "over_limit"])
def test_a_developer_cache_control_is_kept_only_within_the_four_mark_limit(
case: _Case,
marked_system: list[JsonValue],
@ -988,9 +990,7 @@ def test_a_developer_cache_control_is_kept_only_within_the_four_mark_limit(
assert _request_body(route)["system"] == (marked_system if keeps_developer_mark else unmarked_system)
@pytest.mark.parametrize(
("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS]
)
@pytest.mark.parametrize(("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS])
@pytest.mark.parametrize(
("user_marks", "developer_mark", "tool_mark", "keeps_developer_mark"),
[
@ -1025,9 +1025,7 @@ def test_a_developer_cache_control_counts_tool_marks_and_their_ttl(
assert _request_body(route)["system"] == (marked_system if keeps_developer_mark else unmarked_system)
@pytest.mark.parametrize(
("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS]
)
@pytest.mark.parametrize(("case", "marked_system", "unmarked_system"), [_ANTHROPIC_SYSTEMS, _BEDROCK_CONVERSE_SYSTEMS])
@pytest.mark.parametrize(
("developer_mark", "keeps_developer_mark"),
[({"type": "ephemeral"}, True), ({"type": "ephemeral", "ttl": "1h"}, False)],