mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-26 01:12:21 +00:00
71bf9fe0bb
Root cause: handle_exception_on_proxy always maps non-HTTPException/non-ProxyException errors to 500, even when the original exception carries a status_code (e.g. 429 for rate limits). Made-with: Cursor
215 lines
6.7 KiB
Python
215 lines
6.7 KiB
Python
import datetime as real_datetime
|
|
import json
|
|
import os
|
|
import sys
|
|
|
|
import pytest
|
|
from fastapi import HTTPException
|
|
|
|
from litellm.caching.caching import DualCache
|
|
from litellm.proxy._types import ProxyErrorTypes
|
|
from litellm.proxy.utils import ProxyLogging
|
|
|
|
sys.path.insert(
|
|
0, os.path.abspath("../../..")
|
|
) # Adds the parent directory to the system path
|
|
|
|
|
|
from unittest.mock import MagicMock
|
|
|
|
from litellm.proxy.utils import get_custom_url, join_paths
|
|
|
|
|
|
def test_get_custom_url(monkeypatch):
|
|
monkeypatch.setenv("SERVER_ROOT_PATH", "/litellm")
|
|
custom_url = get_custom_url(request_base_url="http://0.0.0.0:4000", route="ui/")
|
|
assert custom_url == "http://0.0.0.0:4000/litellm/ui/"
|
|
|
|
|
|
def test_proxy_only_error_true_for_llm_route():
|
|
proxy_logging_obj = ProxyLogging(user_api_key_cache=DualCache())
|
|
assert proxy_logging_obj._is_proxy_only_llm_api_error(
|
|
original_exception=Exception(),
|
|
error_type=ProxyErrorTypes.auth_error,
|
|
route="/v1/chat/completions",
|
|
)
|
|
|
|
|
|
def test_proxy_only_error_true_for_info_route():
|
|
proxy_logging_obj = ProxyLogging(user_api_key_cache=DualCache())
|
|
assert (
|
|
proxy_logging_obj._is_proxy_only_llm_api_error(
|
|
original_exception=Exception(),
|
|
error_type=ProxyErrorTypes.auth_error,
|
|
route="/key/info",
|
|
)
|
|
is True
|
|
)
|
|
|
|
|
|
def test_proxy_only_error_false_for_non_llm_non_info_route():
|
|
proxy_logging_obj = ProxyLogging(user_api_key_cache=DualCache())
|
|
assert (
|
|
proxy_logging_obj._is_proxy_only_llm_api_error(
|
|
original_exception=Exception(),
|
|
error_type=ProxyErrorTypes.auth_error,
|
|
route="/key/generate",
|
|
)
|
|
is False
|
|
)
|
|
|
|
|
|
def test_proxy_only_error_false_for_other_error_type():
|
|
proxy_logging_obj = ProxyLogging(user_api_key_cache=DualCache())
|
|
assert (
|
|
proxy_logging_obj._is_proxy_only_llm_api_error(
|
|
original_exception=Exception(),
|
|
error_type=None,
|
|
route="/v1/chat/completions",
|
|
)
|
|
is False
|
|
)
|
|
|
|
|
|
def test_get_model_group_info_order():
|
|
from litellm import Router
|
|
from litellm.proxy.proxy_server import _get_model_group_info
|
|
|
|
router = Router(
|
|
model_list=[
|
|
{
|
|
"model_name": "openai/tts-1",
|
|
"litellm_params": {
|
|
"model": "openai/tts-1",
|
|
"api_key": "sk-1234",
|
|
},
|
|
},
|
|
{
|
|
"model_name": "openai/gpt-3.5-turbo",
|
|
"litellm_params": {
|
|
"model": "openai/gpt-3.5-turbo",
|
|
"api_key": "sk-1234",
|
|
},
|
|
},
|
|
]
|
|
)
|
|
model_list = _get_model_group_info(
|
|
llm_router=router,
|
|
all_models_str=["openai/tts-1", "openai/gpt-3.5-turbo"],
|
|
model_group=None,
|
|
)
|
|
|
|
model_groups = [m.model_group for m in model_list]
|
|
assert model_groups == ["openai/tts-1", "openai/gpt-3.5-turbo"]
|
|
|
|
|
|
def test_join_paths_no_duplication():
|
|
"""Test that join_paths doesn't duplicate route when base_path already ends with it"""
|
|
result = join_paths(
|
|
base_path="http://0.0.0.0:4000/my-custom-path/", route="/my-custom-path"
|
|
)
|
|
assert result == "http://0.0.0.0:4000/my-custom-path"
|
|
|
|
|
|
def test_join_paths_normal_join():
|
|
"""Test normal path joining"""
|
|
result = join_paths(base_path="http://0.0.0.0:4000", route="/api/v1")
|
|
assert result == "http://0.0.0.0:4000/api/v1"
|
|
|
|
|
|
def test_join_paths_with_trailing_slash():
|
|
"""Test path joining with trailing slash on base_path"""
|
|
result = join_paths(base_path="http://0.0.0.0:4000/", route="api/v1")
|
|
assert result == "http://0.0.0.0:4000/api/v1"
|
|
|
|
|
|
def test_join_paths_empty_base():
|
|
"""Test path joining with empty base_path"""
|
|
result = join_paths(base_path="", route="api/v1")
|
|
assert result == "/api/v1"
|
|
|
|
|
|
def test_join_paths_empty_route():
|
|
"""Test path joining with empty route"""
|
|
result = join_paths(base_path="http://0.0.0.0:4000", route="")
|
|
assert result == "http://0.0.0.0:4000"
|
|
|
|
|
|
def test_join_paths_both_empty():
|
|
"""Test path joining with both empty"""
|
|
result = join_paths(base_path="", route="")
|
|
assert result == "/"
|
|
|
|
|
|
def test_join_paths_nested_path():
|
|
"""Test path joining with nested paths"""
|
|
result = join_paths(base_path="http://0.0.0.0:4000/v1", route="chat/completions")
|
|
assert result == "http://0.0.0.0:4000/v1/chat/completions"
|
|
|
|
|
|
def _patch_today(monkeypatch, year, month, day):
|
|
class PatchedDate(real_datetime.date):
|
|
@classmethod
|
|
def today(cls):
|
|
return real_datetime.date(year, month, day)
|
|
|
|
monkeypatch.setattr("litellm.proxy.utils.date", PatchedDate)
|
|
|
|
|
|
def test_get_projected_spend_over_limit_day_one(monkeypatch):
|
|
from litellm.proxy.utils import _get_projected_spend_over_limit
|
|
|
|
_patch_today(monkeypatch, 2026, 1, 1)
|
|
result = _get_projected_spend_over_limit(100.0, 1.0)
|
|
|
|
assert result is not None
|
|
projected_spend, projected_exceeded_date = result
|
|
assert projected_spend == 3100.0
|
|
assert projected_exceeded_date == real_datetime.date(2026, 1, 1)
|
|
|
|
|
|
def test_get_projected_spend_over_limit_december(monkeypatch):
|
|
from litellm.proxy.utils import _get_projected_spend_over_limit
|
|
|
|
_patch_today(monkeypatch, 2026, 12, 15)
|
|
result = _get_projected_spend_over_limit(100.0, 1.0)
|
|
|
|
assert result is not None
|
|
projected_spend, projected_exceeded_date = result
|
|
assert projected_spend == pytest.approx(214.28571428571428)
|
|
assert projected_exceeded_date == real_datetime.date(2026, 12, 15)
|
|
|
|
|
|
def test_get_projected_spend_over_limit_includes_current_spend(monkeypatch):
|
|
from litellm.proxy.utils import _get_projected_spend_over_limit
|
|
|
|
_patch_today(monkeypatch, 2026, 4, 11)
|
|
result = _get_projected_spend_over_limit(100.0, 200.0)
|
|
|
|
assert result is not None
|
|
projected_spend, projected_exceeded_date = result
|
|
assert projected_spend == 290.0
|
|
assert projected_exceeded_date == real_datetime.date(2026, 4, 21)
|
|
|
|
|
|
def test_handle_exception_on_proxy_preserves_429_status_code():
|
|
"""Exceptions with status_code=429 should not be mapped to 500."""
|
|
from litellm.proxy.utils import handle_exception_on_proxy
|
|
|
|
class FakeRateLimitError(Exception):
|
|
def __init__(self):
|
|
self.status_code = 429
|
|
self.message = "Rate limit exceeded"
|
|
self.param = None
|
|
super().__init__(self.message)
|
|
|
|
result = handle_exception_on_proxy(FakeRateLimitError())
|
|
assert result.code == "429"
|
|
|
|
|
|
def test_handle_exception_on_proxy_defaults_to_500_without_status_code():
|
|
"""Exceptions without status_code should map to 500."""
|
|
from litellm.proxy.utils import handle_exception_on_proxy
|
|
|
|
result = handle_exception_on_proxy(Exception("unknown error"))
|
|
assert result.code == "500"
|