mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-16 23:41:43 +00:00
6fd988cc63
simple_shuffle logs the selected deployment at INFO whenever a weight set applies, so the fallback group's selection line lands between the fallback notice and the success notice and pushed the notice out of the tail-3 window. Filter it the same way the neighbouring get_available_deployment noise is already filtered.
106 lines
3.6 KiB
Python
106 lines
3.6 KiB
Python
import asyncio
|
|
import os
|
|
import time
|
|
import traceback
|
|
|
|
import pytest
|
|
|
|
|
|
import logging
|
|
|
|
import litellm
|
|
from litellm import Router
|
|
|
|
# this tests debug logs from litellm router and litellm proxy server
|
|
from litellm._logging import verbose_logger, verbose_proxy_logger, verbose_router_logger
|
|
from litellm.llms.custom_httpx.async_client_cleanup import close_litellm_async_clients
|
|
|
|
|
|
# this tests debug logs from litellm router and litellm proxy server
|
|
def test_async_fallbacks(caplog):
|
|
# THIS IS A PROD TEST - DO NOT DELETE THIS. Used for testing if litellm proxy verbose logs are human readable
|
|
litellm.set_verbose = False
|
|
litellm.success_callback = []
|
|
litellm.failure_callback = []
|
|
verbose_router_logger.setLevel(level=logging.INFO)
|
|
verbose_logger.setLevel(logging.CRITICAL + 1)
|
|
verbose_proxy_logger.setLevel(logging.CRITICAL + 1)
|
|
model_list = [
|
|
{
|
|
"model_name": "azure/gpt-3.5-turbo",
|
|
"litellm_params": {
|
|
"model": "azure/gpt-4.1-mini",
|
|
"api_key": os.getenv("AZURE_AI_API_KEY"),
|
|
"api_version": os.getenv("AZURE_API_VERSION"),
|
|
"api_base": os.getenv("AZURE_AI_API_BASE"),
|
|
"mock_response": "Hello world",
|
|
},
|
|
"tpm": 240000,
|
|
"rpm": 1800,
|
|
},
|
|
{
|
|
"model_name": "gpt-3.5-turbo",
|
|
"litellm_params": {
|
|
"model": "gpt-3.5-turbo",
|
|
"api_key": "bad-key",
|
|
},
|
|
"tpm": 1000000,
|
|
"rpm": 9000,
|
|
},
|
|
]
|
|
|
|
router = Router(
|
|
model_list=model_list,
|
|
fallbacks=[{"gpt-3.5-turbo": ["azure/gpt-3.5-turbo"]}],
|
|
num_retries=1,
|
|
)
|
|
|
|
user_message = "Hello, how are you?"
|
|
messages = [{"content": user_message, "role": "user"}]
|
|
|
|
async def _make_request():
|
|
try:
|
|
await router.acompletion(
|
|
model="gpt-3.5-turbo", messages=messages, max_tokens=1
|
|
)
|
|
router.reset()
|
|
except litellm.Timeout:
|
|
pass
|
|
except Exception as e:
|
|
pytest.fail(f"An exception occurred: {e}")
|
|
finally:
|
|
router.reset()
|
|
# Close cached aiohttp/httpx clients before the event loop ends
|
|
# to prevent "Unclosed client session" / "Unclosed connector" warnings.
|
|
await close_litellm_async_clients()
|
|
|
|
asyncio.run(_make_request())
|
|
captured_logs = [rec.message for rec in caplog.records]
|
|
|
|
# on circle ci the captured logs get async cleanup noise from the gc (leaked
|
|
# task warnings, plus aiohttp "Unclosed client session"/"Unclosed connector"
|
|
# warnings from cached clients other router tests evicted) - filter it out
|
|
captured_logs = [
|
|
log
|
|
for log in captured_logs
|
|
if "Task exception was never retrieved" not in log
|
|
and "Task was destroyed but it is pending" not in log
|
|
and "get_available_deployment" not in log
|
|
and "Selected deployment for model" not in log
|
|
and "in the Langfuse queue" not in log
|
|
and "Unclosed client session" not in log
|
|
and "Unclosed connector" not in log
|
|
]
|
|
|
|
print("\n Captured caplog records - ", captured_logs)
|
|
|
|
# Define the expected log messages
|
|
# - error request, falling back notice, success notice
|
|
expected_logs = [
|
|
"Falling back to model_group = azure/gpt-3.5-turbo",
|
|
"litellm.acompletion(model=azure/gpt-4.1-mini)\x1b[32m 200 OK\x1b[0m",
|
|
"Successful fallback b/w models.",
|
|
]
|
|
|
|
# Assert that the captured logs match the expected log messages
|
|
assert captured_logs[-3:] == expected_logs
|