mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-17 09:34:48 +00:00
fix(proxy): answer 422, not 500, when a budget rejects a Responses body whose metadata is a JSON string
This commit is contained in:
parent
b8beb98c6b
commit
edf5ccb61c
3 changed files with 40 additions and 9 deletions
|
|
@ -638,7 +638,7 @@ def _has_tag_budget(tag: "LiteLLM_TagTable | None") -> bool:
|
|||
return tag.litellm_budget_table.max_budget is not None
|
||||
|
||||
|
||||
def _normalized_metadata_slot(
|
||||
def normalized_metadata_slot(
|
||||
request_data: MutableMapping[str, object], metadata_variable_name: str
|
||||
) -> dict[str, object]:
|
||||
"""Return the request's metadata slot as a dict, normalising it in place first.
|
||||
|
|
@ -2407,7 +2407,7 @@ class LiteLLMProxyRequestSetup:
|
|||
return
|
||||
tags: Final = _STRING_TAGS_ADAPTER.validate_python(raw_tags)
|
||||
slot: Final = metadata_key if metadata_key is not None else get_metadata_variable_name_from_kwargs(request_data)
|
||||
metadata: Final = _normalized_metadata_slot(request_data, slot)
|
||||
metadata: Final = normalized_metadata_slot(request_data, slot)
|
||||
existing: Final = metadata.get("tags")
|
||||
if existing and not isinstance(existing, list):
|
||||
return
|
||||
|
|
@ -2441,7 +2441,7 @@ class LiteLLMProxyRequestSetup:
|
|||
caller_tags: Final = caller_metadata.get("tags") if isinstance(caller_metadata, dict) else None
|
||||
if not isinstance(caller_tags, list) or not caller_tags:
|
||||
return
|
||||
metadata: Final = _normalized_metadata_slot(request_data, slot)
|
||||
metadata: Final = normalized_metadata_slot(request_data, slot)
|
||||
existing_tags: Final = metadata.get("tags")
|
||||
metadata["tags"] = LiteLLMProxyRequestSetup.merge_tags(
|
||||
request_tags=existing_tags if isinstance(existing_tags, list) else None,
|
||||
|
|
@ -2463,7 +2463,7 @@ class LiteLLMProxyRequestSetup:
|
|||
return
|
||||
|
||||
_metadata_variable_name: Final = get_metadata_variable_name_from_kwargs(request_data)
|
||||
metadata: Final = _normalized_metadata_slot(request_data, _metadata_variable_name)
|
||||
metadata: Final = normalized_metadata_slot(request_data, _metadata_variable_name)
|
||||
|
||||
existing_tags: Final = metadata.get("tags")
|
||||
metadata["tags"] = LiteLLMProxyRequestSetup.merge_tags(
|
||||
|
|
@ -2515,7 +2515,7 @@ class LiteLLMProxyRequestSetup:
|
|||
# from (litellm_metadata vs metadata) so the merged tags are visible
|
||||
# to _tag_max_budget_check.
|
||||
_metadata_variable_name: Final = get_metadata_variable_name_from_kwargs(request_data)
|
||||
metadata: Final = _normalized_metadata_slot(request_data, _metadata_variable_name)
|
||||
metadata: Final = normalized_metadata_slot(request_data, _metadata_variable_name)
|
||||
|
||||
existing_tags: Final = metadata.get("tags")
|
||||
metadata["tags"] = LiteLLMProxyRequestSetup.merge_tags(
|
||||
|
|
|
|||
|
|
@ -250,7 +250,11 @@ from litellm.proxy.hooks.sensitive_data_routing import ( # noqa: F401, RUF100
|
|||
PROXY_SensitiveDataRoutingHandler,
|
||||
_PROXY_SensitiveDataRoutingHandler, # pyright: ignore[reportPrivateUsage,reportUnusedImport] # backwards-compatible package export
|
||||
)
|
||||
from litellm.proxy.litellm_pre_call_utils import LiteLLMProxyRequestSetup, add_guardrails_from_auth_metadata
|
||||
from litellm.proxy.litellm_pre_call_utils import (
|
||||
LiteLLMProxyRequestSetup,
|
||||
add_guardrails_from_auth_metadata,
|
||||
normalized_metadata_slot,
|
||||
)
|
||||
from litellm.proxy.management_helpers.key_settings_audit import with_settings_updated_at
|
||||
from litellm.proxy.policy_engine.pipeline_executor import (
|
||||
PipelineExecutor,
|
||||
|
|
@ -3966,9 +3970,7 @@ class ProxyLogging:
|
|||
)
|
||||
request_data["litellm_logging_obj"] = litellm_logging_obj # rebind-ok: lifted then popped by the caller
|
||||
request_data.update({REJECTED_BEFORE_REQUEST_SETUP_KEY: REJECTED_BEFORE_REQUEST_SETUP})
|
||||
if "metadata" not in request_data:
|
||||
request_data["metadata"] = {}
|
||||
request_data["metadata"].update(user_api_key_logged_metadata)
|
||||
normalized_metadata_slot(request_data, "metadata").update(user_api_key_logged_metadata)
|
||||
|
||||
if litellm_logging_obj is not None:
|
||||
## UPDATE LOGGING INPUT
|
||||
|
|
|
|||
|
|
@ -4,6 +4,7 @@ and ``_handle_logging_proxy_only_error``."""
|
|||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import json
|
||||
from collections.abc import Callable
|
||||
from datetime import datetime
|
||||
from types import MappingProxyType
|
||||
|
|
@ -1192,3 +1193,31 @@ async def test_handle_logging_proxy_only_path_flags_requests_that_died_before_re
|
|||
blocked_by_guardrail.get(REJECTED_BEFORE_REQUEST_SETUP_KEY) is REJECTED_BEFORE_REQUEST_SETUP,
|
||||
blocked_by_guardrail["litellm_call_id"],
|
||||
) == (True, "client-supplied", False, "call-1")
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_auth_failure_on_a_responses_body_with_a_json_string_metadata_is_logged_against_the_key(
|
||||
proxy_logging, monkeypatch
|
||||
):
|
||||
from litellm.proxy import proxy_server
|
||||
|
||||
recorded: list[dict] = []
|
||||
|
||||
class _RecordingLogger(CustomLogger):
|
||||
async def async_log_failure_event(self, kwargs, response_obj, start_time, end_time):
|
||||
recorded.append(kwargs)
|
||||
|
||||
monkeypatch.setattr(proxy_server, "llm_router", None)
|
||||
monkeypatch.setattr(litellm, "callbacks", [_RecordingLogger()])
|
||||
proxy_logging.alert_types = []
|
||||
|
||||
await proxy_logging.post_call_failure_hook(
|
||||
request_data={"model": "internal-model", "input": "hi", "metadata": json.dumps({"tags": ["checkout"]})},
|
||||
original_exception=litellm.BudgetExceededError(current_cost=1.0, max_budget=0.5),
|
||||
user_api_key_dict=UserAPIKeyAuth(api_key="sk-hashed", key_alias="checkout-key", request_route="/v1/responses"),
|
||||
error_type=ProxyErrorTypes.auth_error,
|
||||
route="/v1/responses",
|
||||
)
|
||||
|
||||
assert len(recorded) == 1
|
||||
assert recorded[0]["standard_logging_object"]["metadata"]["user_api_key_alias"] == "checkout-key"
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue