scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-14 23:21:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 102

fix(responses): warn when spend-log session query hits row cap

Browse source
This commit is contained in:
Devin AI 2026-07-17 15:55:34 +00:00
parent bbb407b657
commit 3b1ebbc7cb
2 changed files with 46 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
litellm/responses/litellm_completion_transformation/session_handler.py
View file

@ -295,6 +295,15 @@ class ResponsesSessionHandler:
query, previous_response_id, DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
)
if len(spend_logs) >= DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION:
verbose_proxy_logger.warning(
"Responses session for previous_response_id=%s hit the %d-row cap "
"(DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION); the oldest turns were dropped. "
"Raise the limit if you need more history.",
previous_response_id,
DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION,
)
verbose_proxy_logger.debug(
"Found the following spend logs for previous response id %s: %s",
previous_response_id,

37
tests/test_litellm/responses/litellm_completion_transformation/test_session_handler.py
View file

@ -491,3 +491,40 @@ async def test_get_all_spend_logs_returns_most_recent_in_ascending_order():
executed_query = mock_prisma_client.db.query_raw.await_args.args[0]
assert 'ORDER BY "endTime" DESC' in executed_query
assert executed_query.rstrip().rstrip(";").endswith('ORDER BY "endTime" ASC')
@pytest.mark.asyncio
async def test_get_all_spend_logs_warns_when_cap_is_hit():
"""
Operators need a runtime signal when a session is truncated: when the query returns
the full cap of rows (older turns silently dropped), a warning must be emitted.
"""
cap = litellm.constants.DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
mock_prisma_client = AsyncMock()
mock_prisma_client.db.query_raw = AsyncMock(return_value=[{} for _ in range(cap)])
with _patched_prisma_client(mock_prisma_client):
with patch.object(session_handler.verbose_proxy_logger, "warning") as mock_warning:
await ResponsesSessionHandler.get_all_spend_logs_for_previous_response_id(
"resp_previous_id"
)
mock_warning.assert_called_once()
@pytest.mark.asyncio
async def test_get_all_spend_logs_does_not_warn_below_cap():
"""
Below the cap nothing is truncated, so no truncation warning should be emitted.
"""
cap = litellm.constants.DEFAULT_MAX_SPEND_LOGS_PER_RESPONSES_SESSION
mock_prisma_client = AsyncMock()
mock_prisma_client.db.query_raw = AsyncMock(return_value=[{} for _ in range(cap - 1)])
with _patched_prisma_client(mock_prisma_client):
with patch.object(session_handler.verbose_proxy_logger, "warning") as mock_warning:
await ResponsesSessionHandler.get_all_spend_logs_for_previous_response_id(
"resp_previous_id"
)
mock_warning.assert_not_called()