mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-17 09:34:48 +00:00
fix(logging): log and bill Anthropic messages compaction responses on the Rust route
The Rust route hands the served Anthropic body to the logger, which validates it against AnthropicResponse. That model had no compaction content block and no compaction stop reason, so a compaction reply raised and wrote no spend row.
This commit is contained in:
parent
cdc280c867
commit
c8097f9c37
2 changed files with 171 additions and 99 deletions
|
|
@ -698,6 +698,14 @@ class AnthropicResponseContentBlockRedactedThinking(LiteLLMBaseModel):
|
|||
data: str
|
||||
|
||||
|
||||
class AnthropicResponseContentBlockCompaction(LiteLLMBaseModel):
|
||||
model_config = ConfigDict(frozen=True)
|
||||
|
||||
type: Literal["compaction"]
|
||||
content: str | None
|
||||
signature: str | None = None
|
||||
|
||||
|
||||
class AnthropicResponseUsageBlock(LiteLLMBaseModel):
|
||||
model_config = ConfigDict(extra="allow")
|
||||
|
||||
|
|
@ -711,7 +719,7 @@ class AnthropicOutputTokensDetails(LiteLLMBaseModel):
|
|||
thinking_tokens: int | None = None
|
||||
|
||||
|
||||
AnthropicFinishReason = Literal["end_turn", "max_tokens", "stop_sequence", "tool_use", "refusal"]
|
||||
AnthropicFinishReason = Literal["end_turn", "max_tokens", "stop_sequence", "tool_use", "refusal", "compaction"]
|
||||
|
||||
|
||||
class AnthropicResponse(LiteLLMBaseModel):
|
||||
|
|
@ -729,6 +737,7 @@ class AnthropicResponse(LiteLLMBaseModel):
|
|||
| AnthropicResponseContentBlockToolUse
|
||||
| AnthropicResponseContentBlockThinking
|
||||
| AnthropicResponseContentBlockRedactedThinking
|
||||
| AnthropicResponseContentBlockCompaction
|
||||
]
|
||||
"""Content generated by the model."""
|
||||
|
||||
|
|
|
|||
|
|
@ -75,6 +75,7 @@ import inspect
|
|||
from typing import List, Optional
|
||||
from litellm.types.utils import LiteLLMCommonStrings
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def logging_obj():
|
||||
return LitellmLogging(
|
||||
|
|
@ -345,8 +346,10 @@ async def test_mcp_native_structured_replacement_must_match_returned_content(
|
|||
)
|
||||
logging_obj.dynamic_success_callbacks = [NativeReplacement()]
|
||||
returned = await logging_obj.async_post_mcp_tool_call_hook(
|
||||
kwargs={"original_response": result}, response_obj=result,
|
||||
start_time=datetime.datetime.now(), end_time=datetime.datetime.now(),
|
||||
kwargs={"original_response": result},
|
||||
response_obj=result,
|
||||
start_time=datetime.datetime.now(),
|
||||
end_time=datetime.datetime.now(),
|
||||
)
|
||||
assert returned is result
|
||||
assert result.content == [TextContent(type="text", text="native-safe" if same_content else "final-safe")]
|
||||
|
|
@ -355,7 +358,6 @@ async def test_mcp_native_structured_replacement_must_match_returned_content(
|
|||
assert "SECRET-1234" not in result.model_dump_json()
|
||||
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
@pytest.mark.parametrize("mode", ["none", "wrapper", "exception"])
|
||||
@pytest.mark.parametrize("structured", [False, True])
|
||||
|
|
@ -373,8 +375,10 @@ async def test_mcp_direct_content_edit_invalidates_stale_structured_data(logging
|
|||
)
|
||||
logging_obj.dynamic_success_callbacks = [DirectRedactor()]
|
||||
returned = await logging_obj.async_post_mcp_tool_call_hook(
|
||||
kwargs={"original_response": result}, response_obj=result,
|
||||
start_time=datetime.datetime.now(), end_time=datetime.datetime.now(),
|
||||
kwargs={"original_response": result},
|
||||
response_obj=result,
|
||||
start_time=datetime.datetime.now(),
|
||||
end_time=datetime.datetime.now(),
|
||||
)
|
||||
assert returned is result
|
||||
assert result.content == [TextContent(type="text", text="[REDACTED]")]
|
||||
|
|
@ -2539,7 +2543,6 @@ async def test_completed_success_selects_one_implemented_hook(
|
|||
assert (thread_id == loop_thread) is (kind != "sync")
|
||||
|
||||
|
||||
|
||||
@pytest.mark.parametrize("for_async_request", [False, True])
|
||||
@pytest.mark.parametrize("private_first", [False, True])
|
||||
@pytest.mark.parametrize("response_kind", ["responses", "completion"])
|
||||
|
|
@ -3227,8 +3230,13 @@ async def test_shadow_snapshot_stays_private_and_is_invalidated_before_logging_g
|
|||
|
||||
class RecordingShadowLogger(ShadowEvalLogger):
|
||||
async def async_log_success_event(
|
||||
self, kwargs: Mapping[str, object], response_obj: object, start_time: object,
|
||||
end_time: object, *, guardrail_snapshot: GuardrailRequestSnapshot | None = None,
|
||||
self,
|
||||
kwargs: Mapping[str, object],
|
||||
response_obj: object,
|
||||
start_time: object,
|
||||
end_time: object,
|
||||
*,
|
||||
guardrail_snapshot: GuardrailRequestSnapshot | None = None,
|
||||
) -> None:
|
||||
shadow_snapshots.append(guardrail_snapshot)
|
||||
await super().async_log_success_event(
|
||||
|
|
@ -3237,13 +3245,20 @@ async def test_shadow_snapshot_stays_private_and_is_invalidated_before_logging_g
|
|||
|
||||
class RecordingLogger(CustomLogger):
|
||||
async def async_log_success_event(
|
||||
self, kwargs: Mapping[str, object], response_obj: object, start_time: object, end_time: object,
|
||||
self,
|
||||
kwargs: Mapping[str, object],
|
||||
response_obj: object,
|
||||
start_time: object,
|
||||
end_time: object,
|
||||
) -> None:
|
||||
other_payloads.append(kwargs)
|
||||
|
||||
class LoggingGuardrail(CustomGuardrail):
|
||||
async def async_logging_hook(
|
||||
self, kwargs: dict[str, object], result: object, call_type: str,
|
||||
self,
|
||||
kwargs: dict[str, object],
|
||||
result: object,
|
||||
call_type: str,
|
||||
) -> tuple[dict[str, object], object]:
|
||||
hook_snapshots.append(logging_obj.shadow_eval_request_snapshot)
|
||||
if hook_mode == "raises":
|
||||
|
|
@ -3255,26 +3270,36 @@ async def test_shadow_snapshot_stays_private_and_is_invalidated_before_logging_g
|
|||
"user_api_key_hash": "test-key",
|
||||
}
|
||||
snapshot: Final = GuardrailRequestSnapshot.capture(
|
||||
{"messages": [{"role": "user", "content": "snapshot-only"}]}, metadata,
|
||||
{"messages": [{"role": "user", "content": "snapshot-only"}]},
|
||||
metadata,
|
||||
)
|
||||
assert snapshot is not None
|
||||
shadow: Final = RecordingShadowLogger(prisma_provider=no_prisma, jobs_cache=InMemoryCache())
|
||||
guardrail: Final = LoggingGuardrail(
|
||||
guardrail_name="late-mask", default_on=True,
|
||||
guardrail_name="late-mask",
|
||||
default_on=True,
|
||||
event_hook=GuardrailEventHooks.pre_call if hook_mode == "disabled" else GuardrailEventHooks.logging_only,
|
||||
)
|
||||
monkeypatch.setattr(litellm, "_async_success_callback", [])
|
||||
logging_obj: Final = LitellmLogging(
|
||||
model="test-model", messages=[], stream=stream, call_type="anthropic_messages",
|
||||
start_time=datetime.datetime.now(), litellm_call_id="private-snapshot", function_id="private-snapshot",
|
||||
model="test-model",
|
||||
messages=[],
|
||||
stream=stream,
|
||||
call_type="anthropic_messages",
|
||||
start_time=datetime.datetime.now(),
|
||||
litellm_call_id="private-snapshot",
|
||||
function_id="private-snapshot",
|
||||
dynamic_async_success_callbacks=[shadow, RecordingLogger(), guardrail],
|
||||
)
|
||||
logging_obj.update_messages([{"role": "user", "content": "logged input"}])
|
||||
logging_obj.update_environment_variables(litellm_params={"metadata": metadata}, optional_params={})
|
||||
logging_obj.shadow_eval_request_snapshot = snapshot
|
||||
payload: Final = {
|
||||
"id": "private-snapshot", "call_type": "anthropic_messages", "metadata": metadata,
|
||||
"model_group": "test-model", "model_parameters": {},
|
||||
"id": "private-snapshot",
|
||||
"call_type": "anthropic_messages",
|
||||
"metadata": metadata,
|
||||
"model_group": "test-model",
|
||||
"model_parameters": {},
|
||||
}
|
||||
|
||||
await logging_obj.async_success_handler(result=ModelResponse(), standard_logging_object=payload)
|
||||
|
|
@ -3731,7 +3756,9 @@ def test_sentry_event_scrubber_initialization(monkeypatch):
|
|||
call_args = mock_init.call_args[1]
|
||||
assert call_args["send_default_pii"] is False
|
||||
assert call_args["event_scrubber"].recursive is True
|
||||
assert {name.lower() for name in SENTRY_PII_DENYLIST} <= {name.lower() for name in call_args["event_scrubber"].denylist}
|
||||
assert {name.lower() for name in SENTRY_PII_DENYLIST} <= {
|
||||
name.lower() for name in call_args["event_scrubber"].denylist
|
||||
}
|
||||
assert call_args["before_send"] is call_args["before_send_transaction"]
|
||||
|
||||
|
||||
|
|
@ -3748,7 +3775,9 @@ def test_sentry_send_default_pii_opt_in(monkeypatch):
|
|||
|
||||
call_args = mock_init.call_args[1]
|
||||
assert call_args["send_default_pii"] is True
|
||||
assert not {name.lower() for name in SENTRY_PII_DENYLIST} & {name.lower() for name in call_args["event_scrubber"].denylist}
|
||||
assert not {name.lower() for name in SENTRY_PII_DENYLIST} & {
|
||||
name.lower() for name in call_args["event_scrubber"].denylist
|
||||
}
|
||||
|
||||
|
||||
def test_get_masked_values():
|
||||
|
|
@ -5018,12 +5047,10 @@ def test_reported_cost_respects_configured_pricing_and_additional_charges(
|
|||
else {}
|
||||
),
|
||||
}
|
||||
}
|
||||
}
|
||||
)
|
||||
logging_obj.model_call_details.update(
|
||||
{"additional_response_cost": additional_cost, "cache_hit": cache_hit}
|
||||
},
|
||||
},
|
||||
)
|
||||
logging_obj.model_call_details.update({"additional_response_cost": additional_cost, "cache_hit": cache_hit})
|
||||
logging_obj._provider_reported_cost = reported_cost
|
||||
response: Final = ModelResponse(
|
||||
model=pricing_model,
|
||||
|
|
@ -6512,6 +6539,31 @@ def test_handle_anthropic_messages_response_logging_reparses_the_served_body_wit
|
|||
assert logged.usage.completion_tokens == 7 # type: ignore[attr-defined]
|
||||
|
||||
|
||||
def test_handle_anthropic_messages_response_logging_reparses_a_served_compaction_response():
|
||||
logging_obj = _anthropic_messages_logging_obj()
|
||||
logging_obj.optional_params = {}
|
||||
compaction_block = {"type": "compaction", "content": "Summary of the chat so far.", "signature": "sig-abc"}
|
||||
|
||||
logged = logging_obj._handle_anthropic_messages_response_logging(
|
||||
result={
|
||||
"id": "msg_compacted",
|
||||
"type": "message",
|
||||
"role": "assistant",
|
||||
"model": "claude-sonnet-5-5",
|
||||
"content": [compaction_block],
|
||||
"stop_reason": "compaction",
|
||||
"stop_sequence": None,
|
||||
"usage": {"input_tokens": 410, "output_tokens": 95},
|
||||
}
|
||||
)
|
||||
|
||||
assert isinstance(logged, ModelResponse)
|
||||
assert logged.id == "msg_compacted"
|
||||
assert logged.usage.prompt_tokens == 410 # type: ignore[attr-defined]
|
||||
assert logged.usage.completion_tokens == 95 # type: ignore[attr-defined]
|
||||
assert logged.choices[0].message.provider_specific_fields["compaction_blocks"] == [compaction_block] # type: ignore[union-attr]
|
||||
|
||||
|
||||
def test_handle_anthropic_messages_response_logging_passes_model_response_through():
|
||||
"""Anthropic-native path already yields a ModelResponse; it must be returned unchanged."""
|
||||
logging_obj = _anthropic_messages_logging_obj()
|
||||
|
|
@ -6950,8 +7002,11 @@ async def test_background_interaction_completion_rebills_after_in_progress_succe
|
|||
|
||||
class SyncRecorder(CustomLogger):
|
||||
def log_success_event(
|
||||
self, kwargs: dict[str, object], response_obj: object,
|
||||
start_time: dt.datetime, end_time: dt.datetime,
|
||||
self,
|
||||
kwargs: dict[str, object],
|
||||
response_obj: object,
|
||||
start_time: dt.datetime,
|
||||
end_time: dt.datetime,
|
||||
) -> None:
|
||||
events.put((response_obj, kwargs.get("response_cost")))
|
||||
logged.set()
|
||||
|
|
@ -6984,11 +7039,11 @@ async def test_background_interaction_completion_rebills_after_in_progress_succe
|
|||
assert logging_obj.model_call_details["response_cost"] > 0
|
||||
assert logging_obj.model_call_details["standard_logging_object"]["total_tokens"] == 175
|
||||
|
||||
|
||||
assert events.qsize() == 2
|
||||
assert events.get_nowait() == (in_progress, None)
|
||||
assert events.get_nowait() == (completed, logging_obj.model_call_details["response_cost"])
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_background_interaction_completion_prices_the_settled_body_itself():
|
||||
"""
|
||||
|
|
@ -8022,7 +8077,9 @@ def test_debugging_log_with_json_logs_tolerates_missing_headers(logging_obj, mon
|
|||
logging_obj.litellm_request_debug = True
|
||||
|
||||
with patch("litellm.litellm_core_utils.litellm_logging.verbose_logger.warning") as warning:
|
||||
logging_obj._print_llm_call_debugging_log(api_base="https://api.openai.com/v1", headers=None, additional_args={})
|
||||
logging_obj._print_llm_call_debugging_log(
|
||||
api_base="https://api.openai.com/v1", headers=None, additional_args={}
|
||||
)
|
||||
|
||||
assert "https://api.openai.com/v1" in warning.call_args.kwargs["extra"]["api_base"]
|
||||
|
||||
|
|
@ -10159,8 +10216,6 @@ def test_get_assembled_streaming_response_bills_a_provider_reported_usage_cost()
|
|||
assert logging_obj.response_cost_calculator(result=assembled) == 0.0042
|
||||
|
||||
|
||||
|
||||
|
||||
def test_response_cost_calculator_prices_terminal_responses_event_from_its_response():
|
||||
logging_obj: Final = _responses_stream_logging_obj()
|
||||
inner_response: Final = ResponsesAPIResponse(
|
||||
|
|
@ -10390,9 +10445,7 @@ def test_standard_logging_retries():
|
|||
]
|
||||
)
|
||||
|
||||
with patch.object(
|
||||
customHandler, "log_failure_event", new=MagicMock()
|
||||
) as mock_client:
|
||||
with patch.object(customHandler, "log_failure_event", new=MagicMock()) as mock_client:
|
||||
try:
|
||||
router.completion(
|
||||
model="gpt-3.5-turbo",
|
||||
|
|
@ -10404,19 +10457,10 @@ def test_standard_logging_retries():
|
|||
pass
|
||||
|
||||
assert mock_client.call_count == 2
|
||||
assert mock_client.call_args_list[0].kwargs["kwargs"]["standard_logging_object"]["trace_id"] is not None
|
||||
assert (
|
||||
mock_client.call_args_list[0].kwargs["kwargs"]["standard_logging_object"][
|
||||
"trace_id"
|
||||
]
|
||||
is not None
|
||||
)
|
||||
assert (
|
||||
mock_client.call_args_list[0].kwargs["kwargs"]["standard_logging_object"][
|
||||
"trace_id"
|
||||
]
|
||||
== mock_client.call_args_list[1].kwargs["kwargs"][
|
||||
"standard_logging_object"
|
||||
]["trace_id"]
|
||||
mock_client.call_args_list[0].kwargs["kwargs"]["standard_logging_object"]["trace_id"]
|
||||
== mock_client.call_args_list[1].kwargs["kwargs"]["standard_logging_object"]["trace_id"]
|
||||
)
|
||||
|
||||
|
||||
|
|
@ -10683,9 +10727,7 @@ def test_signoz_dispatch_requires_an_endpoint(monkeypatch):
|
|||
custom_logger_init_args={},
|
||||
)
|
||||
assert created is None
|
||||
assert not [
|
||||
cb for cb in logging_module._in_memory_loggers if getattr(cb, "callback_name", None) == "signoz"
|
||||
]
|
||||
assert not [cb for cb in logging_module._in_memory_loggers if getattr(cb, "callback_name", None) == "signoz"]
|
||||
finally:
|
||||
logging_module._in_memory_loggers.clear()
|
||||
monkeypatch.delenv("LITELLM_OTEL_V2", raising=False)
|
||||
|
|
@ -10827,7 +10869,11 @@ def test_standard_logging_payload_reports_the_deployment_entry_the_request_was_b
|
|||
}
|
||||
}
|
||||
logging_obj.update_environment_variables(
|
||||
model="gpt-5.4-nano", user="", optional_params={}, litellm_params=litellm_params, custom_llm_provider="openai"
|
||||
model="gpt-5.4-nano",
|
||||
user="",
|
||||
optional_params={},
|
||||
litellm_params=litellm_params,
|
||||
custom_llm_provider="openai",
|
||||
)
|
||||
response: Final = ModelResponse(
|
||||
model="gpt-5.4-nano",
|
||||
|
|
@ -11139,10 +11185,7 @@ def test_enterprise_alerting_loggers_falls_back_without_disabling_callback_contr
|
|||
)
|
||||
logging_module._enterprise_alerting_loggers.cache_clear()
|
||||
for module_name in tuple(sys.modules):
|
||||
if any(
|
||||
module_name == prefix or module_name.startswith(f"{prefix}.")
|
||||
for prefix in module_prefixes
|
||||
):
|
||||
if any(module_name == prefix or module_name.startswith(f"{prefix}.") for prefix in module_prefixes):
|
||||
monkeypatch.delitem(sys.modules, module_name, raising=False)
|
||||
monkeypatch.setattr(sys, "meta_path", [BlockPagerDutyFinder(), *sys.meta_path])
|
||||
|
||||
|
|
@ -11262,13 +11305,16 @@ def _vcr_outcome_gate(request, vcr):
|
|||
yield
|
||||
record_vcr_outcome(request, vcr)
|
||||
|
||||
|
||||
@pytest_asyncio.fixture(loop_scope="function")
|
||||
async def drain_logging_worker(isolate_litellm_state: None) -> AsyncIterator[None]:
|
||||
yield
|
||||
await asyncio.wait_for(GLOBAL_LOGGING_WORKER.flush(), timeout=LOGGING_WORKER_DRAIN_TIMEOUT_SECONDS)
|
||||
|
||||
|
||||
LOGGING_WORKER_DRAIN_TIMEOUT_SECONDS: Final = LOGGING_WORKER_MAX_TIME_PER_COROUTINE + 5.0
|
||||
|
||||
|
||||
@pytest.fixture(scope="function")
|
||||
def isolate_litellm_state():
|
||||
"""
|
||||
|
|
@ -11306,6 +11352,7 @@ def isolate_litellm_state():
|
|||
if attr in _DEFAULTS:
|
||||
setattr(litellm, attr, _DEFAULTS[attr])
|
||||
|
||||
|
||||
_LIST_ATTRS = (
|
||||
"callbacks",
|
||||
"success_callback",
|
||||
|
|
@ -11333,6 +11380,7 @@ _SCALAR_ATTRS = (
|
|||
|
||||
_DEFAULTS: dict = {}
|
||||
|
||||
|
||||
@pytest.fixture(scope="module")
|
||||
def setup_and_teardown():
|
||||
"""
|
||||
|
|
@ -11355,6 +11403,7 @@ def setup_and_teardown():
|
|||
litellm.in_memory_llm_clients_cache.flush_cache()
|
||||
yield
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
@pytest.mark.parametrize(
|
||||
"response_obj,expected_values",
|
||||
|
|
@ -11402,6 +11451,7 @@ def test_get_usage(response_obj, expected_values):
|
|||
assert usage.completion_tokens == expected_values[1]
|
||||
assert usage.total_tokens == expected_values[2]
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_usage_from_image_generation_response():
|
||||
"""
|
||||
|
|
@ -11449,6 +11499,7 @@ def test_get_usage_from_image_generation_response():
|
|||
assert usage.completion_tokens_details.image_tokens == 272
|
||||
assert usage.completion_tokens_details.text_tokens == 100
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_additional_headers():
|
||||
additional_headers = {
|
||||
|
|
@ -11489,10 +11540,12 @@ def test_get_additional_headers():
|
|||
assert additional_logging_headers.get("llm_provider-request-id") == "req_01F6CycZZPSHKRCCctcS1Vto"
|
||||
assert additional_logging_headers.get("llm_provider-anthropic-ratelimit-requests-reset") == "2024-10-29T23:57:40Z"
|
||||
|
||||
|
||||
def all_fields_present(standard_logging_metadata: StandardLoggingMetadata):
|
||||
for field in StandardLoggingMetadata.__annotations__.keys():
|
||||
assert field in standard_logging_metadata
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
@pytest.mark.parametrize(
|
||||
"metadata_key, metadata_value",
|
||||
|
|
@ -11524,6 +11577,7 @@ def test_get_standard_logging_metadata(metadata_key, metadata_value):
|
|||
# Assert that the specific metadata field is set correctly
|
||||
assert standard_logging_metadata[metadata_key] == metadata_value
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_standard_logging_metadata_user_api_key_hash():
|
||||
valid_hash = "a" * 64 # 64 character string
|
||||
|
|
@ -11531,6 +11585,7 @@ def test_get_standard_logging_metadata_user_api_key_hash():
|
|||
result = StandardLoggingPayloadSetup.get_standard_logging_metadata(metadata)
|
||||
assert result["user_api_key_hash"] == valid_hash
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_standard_logging_metadata_invalid_user_api_key():
|
||||
invalid_hash = "not_a_valid_hash"
|
||||
|
|
@ -11539,6 +11594,7 @@ def test_get_standard_logging_metadata_invalid_user_api_key():
|
|||
all_fields_present(result)
|
||||
assert result["user_api_key_hash"] is None
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_standard_logging_metadata_non_string_user_api_key():
|
||||
"""Non-string user_api_key should not be set as user_api_key_hash."""
|
||||
|
|
@ -11547,6 +11603,7 @@ def test_get_standard_logging_metadata_non_string_user_api_key():
|
|||
all_fields_present(result)
|
||||
assert result["user_api_key_hash"] is None
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_standard_logging_metadata_none_user_api_key():
|
||||
"""None user_api_key should not be set as user_api_key_hash."""
|
||||
|
|
@ -11555,6 +11612,7 @@ def test_get_standard_logging_metadata_none_user_api_key():
|
|||
all_fields_present(result)
|
||||
assert result["user_api_key_hash"] is None
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_standard_logging_metadata_invalid_keys():
|
||||
metadata = {
|
||||
|
|
@ -11568,6 +11626,7 @@ def test_get_standard_logging_metadata_invalid_keys():
|
|||
assert "invalid_key" not in result
|
||||
assert "another_invalid_key" not in result
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_cleanup_timestamps():
|
||||
"""Test cleanup_timestamps with different input types"""
|
||||
|
|
@ -11602,6 +11661,7 @@ def test_cleanup_timestamps():
|
|||
with pytest.raises(ValueError, match="start_time is required, got=invalid of type <class 'str'>"):
|
||||
StandardLoggingPayloadSetup.cleanup_timestamps("invalid", end_float, completion_float)
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_model_cost_information():
|
||||
"""Test get_model_cost_information with different inputs"""
|
||||
|
|
@ -11634,6 +11694,7 @@ def test_get_model_cost_information():
|
|||
# assert all fields in StandardLoggingModelInformation are present
|
||||
assert all(field in result for field in StandardLoggingModelInformation.__annotations__)
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_model_cost_information_custom_pricing_uses_base_model():
|
||||
result = StandardLoggingPayloadSetup.get_model_cost_information(
|
||||
|
|
@ -11645,6 +11706,7 @@ def test_get_model_cost_information_custom_pricing_uses_base_model():
|
|||
assert result["model_map_value"] is not None
|
||||
assert result["model_map_key"] != "invoke_test_claude"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_standard_logging_payload_uses_deployment_when_no_base_model():
|
||||
"""metadata["deployment"] is used for cost-map lookup when base_model is not set."""
|
||||
|
|
@ -11702,6 +11764,7 @@ def test_standard_logging_payload_uses_deployment_when_no_base_model():
|
|||
assert payload["model_map_information"]["model_map_value"] is not None
|
||||
assert payload["model_map_information"]["model_map_key"] != "invoke_test_claude"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_hidden_params():
|
||||
"""Test get_hidden_params with different inputs"""
|
||||
|
|
@ -11737,6 +11800,7 @@ def test_get_hidden_params():
|
|||
# assert all fields in StandardLoggingHiddenParams are present
|
||||
assert all(field in result for field in StandardLoggingHiddenParams.__annotations__)
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_final_response_obj():
|
||||
"""Test get_final_response_obj with different input types and redaction scenarios"""
|
||||
|
|
@ -11768,6 +11832,7 @@ def test_get_final_response_obj():
|
|||
# Reset litellm.turn_off_message_logging to its original value
|
||||
litellm.turn_off_message_logging = False
|
||||
|
||||
|
||||
def testget_standard_logging_payload_trace_id():
|
||||
"""Test get_standard_logging_payload_trace_id with different input scenarios"""
|
||||
# Test case 1: When litellm_trace_id is provided in litellm_params
|
||||
|
|
@ -11805,6 +11870,7 @@ def testget_standard_logging_payload_trace_id():
|
|||
assert result == "12345"
|
||||
assert isinstance(result, str)
|
||||
|
||||
|
||||
def testget_standard_logging_payload_trace_id_prioritizes_trace_id_when_flag_on(monkeypatch):
|
||||
"""With request_correlation_in_logs on, an explicit litellm_trace_id wins over litellm_session_id."""
|
||||
from unittest.mock import MagicMock
|
||||
|
|
@ -11819,6 +11885,7 @@ def testget_standard_logging_payload_trace_id_prioritizes_trace_id_when_flag_on(
|
|||
)
|
||||
assert result == "the-trace-id"
|
||||
|
||||
|
||||
def testget_standard_logging_payload_trace_id_prioritizes_session_id_when_flag_off(monkeypatch):
|
||||
"""With request_correlation_in_logs off (default), legacy behavior is preserved:
|
||||
litellm_session_id still wins over litellm_trace_id."""
|
||||
|
|
@ -11834,6 +11901,7 @@ def testget_standard_logging_payload_trace_id_prioritizes_session_id_when_flag_o
|
|||
)
|
||||
assert result == "the-session-id"
|
||||
|
||||
|
||||
def testget_standard_logging_payload_session_id_when_flag_on(monkeypatch):
|
||||
"""Test get_standard_logging_payload_session_id with different input scenarios, flag enabled"""
|
||||
from unittest.mock import MagicMock
|
||||
|
|
@ -11886,6 +11954,7 @@ def testget_standard_logging_payload_session_id_when_flag_on(monkeypatch):
|
|||
)
|
||||
assert result == ""
|
||||
|
||||
|
||||
def testget_standard_logging_payload_session_id_empty_when_flag_off(monkeypatch):
|
||||
"""When request_correlation_in_logs is off (default), session_id is always empty,
|
||||
even if litellm_session_id was explicitly supplied - preserves the pre-existing
|
||||
|
|
@ -11902,6 +11971,7 @@ def testget_standard_logging_payload_session_id_empty_when_flag_off(monkeypatch)
|
|||
)
|
||||
assert result == ""
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def restore_verbose_logger_level() -> Iterator[None]:
|
||||
original_level: Final = verbose_logger.level
|
||||
|
|
@ -11944,6 +12014,7 @@ def test_truncate_standard_logging_payload():
|
|||
assert len(str(truncated["response"])) < 10_500
|
||||
assert len(str(truncated["error_str"])) < 10_500
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_truncate_standard_logging_payload_keeps_a_partial_payload_intact():
|
||||
"""A payload built with only some of its fields comes back with exactly those keys and values"""
|
||||
|
|
@ -11952,12 +12023,14 @@ def test_truncate_standard_logging_payload_keeps_a_partial_payload_intact():
|
|||
|
||||
assert _custom_logger.truncate_standard_logging_payload_content(partial_payload) == partial_payload
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_strip_trailing_slash():
|
||||
common_api_base = "https://api.test.com"
|
||||
assert StandardLoggingPayloadSetup.strip_trailing_slash(common_api_base + "/") == common_api_base
|
||||
assert StandardLoggingPayloadSetup.strip_trailing_slash(common_api_base) == common_api_base
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_error_information():
|
||||
"""Test get_error_information with different types of exceptions"""
|
||||
|
|
@ -11994,6 +12067,7 @@ def test_get_error_information():
|
|||
assert result["llm_provider"] == "openai"
|
||||
assert result["error_message"] == "litellm.RateLimitError: Test error"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_get_response_time():
|
||||
"""Test get_response_time with different streaming scenarios"""
|
||||
|
|
@ -12029,6 +12103,7 @@ def test_get_response_time():
|
|||
# For streaming, should return completion_start_time - start_time
|
||||
assert response_time == 2.0
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
@pytest.mark.parametrize(
|
||||
"metadata, expected_requester_metadata",
|
||||
|
|
@ -12051,6 +12126,7 @@ def test_standard_logging_metadata_requester_metadata(metadata, expected_request
|
|||
result = StandardLoggingPayloadSetup.get_standard_logging_metadata(metadata)
|
||||
assert result["requester_metadata"] == expected_requester_metadata
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_cost_breakdown_in_standard_logging_payload():
|
||||
"""
|
||||
|
|
@ -12140,6 +12216,7 @@ def test_cost_breakdown_in_standard_logging_payload():
|
|||
|
||||
print("✅ Cost breakdown test passed!")
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_cost_breakdown_missing_in_standard_logging_payload():
|
||||
"""
|
||||
|
|
@ -12200,6 +12277,7 @@ def test_cost_breakdown_missing_in_standard_logging_payload():
|
|||
|
||||
print("✅ Cost breakdown missing test passed!")
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
@pytest.mark.parametrize(
|
||||
"use_combined_usage_object",
|
||||
|
|
@ -12287,6 +12365,7 @@ def test_usage_dict_roundtrip_in_payload(use_combined_usage_object):
|
|||
assert usage_obj["completion_tokens"] == 58
|
||||
assert usage_obj["total_tokens"] == 100
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_standard_logging_payload_uses_actual_model_for_azure_router():
|
||||
from litellm.litellm_core_utils.litellm_logging import (
|
||||
|
|
@ -12335,6 +12414,7 @@ def test_standard_logging_payload_uses_actual_model_for_azure_router():
|
|||
assert payload is not None
|
||||
assert payload["model"] == "azure_ai/gpt-5-nano-2025-08-07"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_standard_logging_payload_uses_actual_model_for_azure_router_with_underscore():
|
||||
from litellm.litellm_core_utils.litellm_logging import (
|
||||
|
|
@ -12383,6 +12463,7 @@ def test_standard_logging_payload_uses_actual_model_for_azure_router_with_unders
|
|||
assert payload is not None
|
||||
assert payload["model"] == "azure_ai/gpt-5-nano-2025-08-07"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_merge_litellm_metadata_basic():
|
||||
"""
|
||||
|
|
@ -12414,6 +12495,7 @@ def test_merge_litellm_metadata_basic():
|
|||
assert result["model_info"] == {"id": "model-123"}
|
||||
assert result["tags"] == ["tag1", "tag2"]
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_merge_litellm_metadata_precedence():
|
||||
"""
|
||||
|
|
@ -12440,6 +12522,7 @@ def test_merge_litellm_metadata_precedence():
|
|||
# litellm_metadata values should only be included if not in metadata
|
||||
assert result["model_group"] == "gpt-4-group"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_merge_litellm_metadata_skip_non_serializable():
|
||||
"""
|
||||
|
|
@ -12472,6 +12555,7 @@ def test_merge_litellm_metadata_skip_non_serializable():
|
|||
assert result["safe_field"] == "safe_value"
|
||||
assert result["model_group"] == "gpt-4-group"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_merge_litellm_metadata_empty_params():
|
||||
"""
|
||||
|
|
@ -12507,6 +12591,7 @@ def test_merge_litellm_metadata_empty_params():
|
|||
result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params)
|
||||
assert result == {}
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_merge_litellm_metadata_bedrock_passthrough_scenario():
|
||||
"""
|
||||
|
|
@ -12566,8 +12651,10 @@ def test_merge_litellm_metadata_bedrock_passthrough_scenario():
|
|||
# Verify total number of fields (9 user fields + 4 model fields = 13)
|
||||
assert len(result) == 13
|
||||
|
||||
|
||||
service_logger = ServiceLogging()
|
||||
|
||||
|
||||
def setup_logging():
|
||||
return Logging(
|
||||
model="gpt-5.5",
|
||||
|
|
@ -12657,6 +12744,7 @@ def test_get_callback_name():
|
|||
# Test string callback
|
||||
assert logging._get_callback_name("callback_string") == "callback_string"
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_is_internal_litellm_proxy_callback():
|
||||
"""
|
||||
|
|
@ -12677,6 +12765,7 @@ def test_is_internal_litellm_proxy_callback():
|
|||
# Test string callback
|
||||
assert logging._is_internal_litellm_proxy_callback("callback_string") == False
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_should_run_sync_callbacks_for_async_calls():
|
||||
"""
|
||||
|
|
@ -12702,6 +12791,7 @@ def test_should_run_sync_callbacks_for_async_calls():
|
|||
litellm.success_callback = [PROXY_MaxIterationsHandler]
|
||||
assert logging._should_run_sync_callbacks_for_async_calls() == False
|
||||
|
||||
|
||||
@pytest.mark.usefixtures("_vcr_outcome_gate", "drain_logging_worker", "isolate_litellm_state", "setup_and_teardown")
|
||||
def test_remove_internal_litellm_callbacks():
|
||||
logging = setup_logging()
|
||||
|
|
@ -12723,6 +12813,7 @@ def test_remove_internal_litellm_callbacks():
|
|||
assert PROXY_MaxIterationsHandler not in filtered
|
||||
assert PROXY_CacheControlCheck not in filtered
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_background_interaction_completion_logs_while_in_progress_handler_is_parked(monkeypatch):
|
||||
"""
|
||||
|
|
@ -12825,14 +12916,9 @@ class CompletionCustomHandler(
|
|||
assert metadata_value is None or isinstance(metadata_value, dict)
|
||||
if metadata_value is not None:
|
||||
if litellm.turn_off_message_logging is True:
|
||||
assert (
|
||||
metadata_value["raw_request"]
|
||||
is LiteLLMCommonStrings.redacted_by_litellm.value
|
||||
)
|
||||
assert metadata_value["raw_request"] is LiteLLMCommonStrings.redacted_by_litellm.value
|
||||
else:
|
||||
assert "raw_request" not in metadata_value or isinstance(
|
||||
metadata_value["raw_request"], str
|
||||
)
|
||||
assert "raw_request" not in metadata_value or isinstance(metadata_value["raw_request"], str)
|
||||
except Exception:
|
||||
print(f"Assertion Error: {traceback.format_exc()}")
|
||||
self.errors.append(traceback.format_exc())
|
||||
|
|
@ -12881,23 +12967,18 @@ class CompletionCustomHandler(
|
|||
assert isinstance(response_obj, litellm.ModelResponseStream)
|
||||
## KWARGS
|
||||
assert isinstance(kwargs["model"], str)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(
|
||||
kwargs["messages"][0], dict
|
||||
)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(kwargs["messages"][0], dict)
|
||||
assert isinstance(kwargs["optional_params"], dict)
|
||||
assert isinstance(kwargs["litellm_params"], dict)
|
||||
assert isinstance(kwargs["start_time"], (datetime, type(None)))
|
||||
assert isinstance(kwargs["stream"], bool)
|
||||
assert isinstance(kwargs["user"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(kwargs["input"], list)
|
||||
and isinstance(kwargs["input"][0], dict)
|
||||
) or isinstance(kwargs["input"], (dict, str))
|
||||
assert (isinstance(kwargs["input"], list) and isinstance(kwargs["input"][0], dict)) or isinstance(
|
||||
kwargs["input"], (dict, str)
|
||||
)
|
||||
assert isinstance(kwargs["api_key"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(
|
||||
kwargs["original_response"], (str, litellm.CustomStreamWrapper)
|
||||
)
|
||||
isinstance(kwargs["original_response"], (str, litellm.CustomStreamWrapper))
|
||||
or inspect.isasyncgen(kwargs["original_response"])
|
||||
or inspect.iscoroutine(kwargs["original_response"])
|
||||
)
|
||||
|
|
@ -12930,9 +13011,7 @@ class CompletionCustomHandler(
|
|||
)
|
||||
## KWARGS
|
||||
assert isinstance(kwargs["model"], str)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(
|
||||
kwargs["messages"][0], dict
|
||||
)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(kwargs["messages"][0], dict)
|
||||
assert isinstance(kwargs["optional_params"], dict)
|
||||
assert isinstance(kwargs["litellm_params"], dict)
|
||||
assert isinstance(kwargs["litellm_params"]["api_base"], str)
|
||||
|
|
@ -12942,10 +13021,7 @@ class CompletionCustomHandler(
|
|||
assert isinstance(kwargs["user"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(kwargs["input"], list)
|
||||
and (
|
||||
isinstance(kwargs["input"][0], dict)
|
||||
or isinstance(kwargs["input"][0], str)
|
||||
)
|
||||
and (isinstance(kwargs["input"][0], dict) or isinstance(kwargs["input"][0], str))
|
||||
) or isinstance(kwargs["input"], (dict, str))
|
||||
assert isinstance(kwargs["api_key"], (str, type(None)))
|
||||
assert isinstance(
|
||||
|
|
@ -12973,9 +13049,7 @@ class CompletionCustomHandler(
|
|||
assert response_obj == None
|
||||
## KWARGS
|
||||
assert isinstance(kwargs["model"], str)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(
|
||||
kwargs["messages"][0], dict
|
||||
)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(kwargs["messages"][0], dict)
|
||||
|
||||
assert isinstance(kwargs["optional_params"], dict)
|
||||
assert isinstance(kwargs["litellm_params"], dict)
|
||||
|
|
@ -12983,15 +13057,12 @@ class CompletionCustomHandler(
|
|||
assert isinstance(kwargs["start_time"], (datetime, type(None)))
|
||||
assert isinstance(kwargs["stream"], bool)
|
||||
assert isinstance(kwargs["user"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(kwargs["input"], list)
|
||||
and isinstance(kwargs["input"][0], dict)
|
||||
) or isinstance(kwargs["input"], (dict, str))
|
||||
assert (isinstance(kwargs["input"], list) and isinstance(kwargs["input"][0], dict)) or isinstance(
|
||||
kwargs["input"], (dict, str)
|
||||
)
|
||||
assert isinstance(kwargs["api_key"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(
|
||||
kwargs["original_response"], (str, litellm.CustomStreamWrapper)
|
||||
)
|
||||
isinstance(kwargs["original_response"], (str, litellm.CustomStreamWrapper))
|
||||
or kwargs["original_response"] == None
|
||||
)
|
||||
assert isinstance(kwargs["additional_args"], (dict, type(None)))
|
||||
|
|
@ -13009,9 +13080,7 @@ class CompletionCustomHandler(
|
|||
assert isinstance(messages, list) and isinstance(messages[0], dict)
|
||||
## KWARGS
|
||||
assert isinstance(kwargs["model"], str)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(
|
||||
kwargs["messages"][0], dict
|
||||
)
|
||||
assert isinstance(kwargs["messages"], list) and isinstance(kwargs["messages"][0], dict)
|
||||
assert isinstance(kwargs["optional_params"], dict)
|
||||
assert isinstance(kwargs["litellm_params"], dict)
|
||||
assert isinstance(kwargs["start_time"], (datetime, type(None)))
|
||||
|
|
@ -13023,9 +13092,7 @@ class CompletionCustomHandler(
|
|||
|
||||
async def async_log_success_event(self, kwargs, response_obj, start_time, end_time):
|
||||
try:
|
||||
print(
|
||||
"in async_log_success_event", kwargs, response_obj, start_time, end_time
|
||||
)
|
||||
print("in async_log_success_event", kwargs, response_obj, start_time, end_time)
|
||||
self.states.append("async_success")
|
||||
## START TIME
|
||||
assert isinstance(start_time, datetime)
|
||||
|
|
@ -13054,9 +13121,7 @@ class CompletionCustomHandler(
|
|||
assert isinstance(kwargs["input"], (list, dict, str))
|
||||
assert isinstance(kwargs["api_key"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(
|
||||
kwargs["original_response"], (str, litellm.CustomStreamWrapper)
|
||||
)
|
||||
isinstance(kwargs["original_response"], (str, litellm.CustomStreamWrapper))
|
||||
or inspect.isasyncgen(kwargs["original_response"])
|
||||
or inspect.iscoroutine(kwargs["original_response"])
|
||||
)
|
||||
|
|
@ -13088,9 +13153,7 @@ class CompletionCustomHandler(
|
|||
assert isinstance(kwargs["input"], (list, str, dict))
|
||||
assert isinstance(kwargs["api_key"], (str, type(None)))
|
||||
assert (
|
||||
isinstance(
|
||||
kwargs["original_response"], (str, litellm.CustomStreamWrapper)
|
||||
)
|
||||
isinstance(kwargs["original_response"], (str, litellm.CustomStreamWrapper))
|
||||
or inspect.isasyncgen(kwargs["original_response"])
|
||||
or inspect.iscoroutine(kwargs["original_response"])
|
||||
or kwargs["original_response"] == None
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue