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(vector-stores): report search failures on the responses API surface too

Browse source
This commit is contained in:
mateo-berri 2026-09-03 01:54:09 -07:00
parent e464e749e0
commit dc28f0eb4b
2 changed files with 64 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
litellm/integrations/vector_store_integrations/vector_store_pre_call_hook.py
View file

@ -17,7 +17,11 @@ import litellm.vector_stores
from litellm._logging import verbose_logger
from litellm.exceptions import VectorStoreSearchError
from litellm.integrations.custom_logger import CustomLogger
from litellm.types.llms.openai import AllMessageValues, ChatCompletionUserMessage
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionUserMessage,
ResponsesAPIResponse,
)
from litellm.types.prompts.init_prompts import PromptSpec
from litellm.types.utils import CallTypes, StandardCallbackDynamicParams
from litellm.types.vector_stores import (
@ -372,6 +376,11 @@ class VectorStorePreCallHook(CustomLogger):
verbose_logger.debug("No search results or search failures found")
return None
if isinstance(response, ResponsesAPIResponse):
if search_failures:
setattr(response, SEARCH_FAILURES_FIELD, list(search_failures))
return response
# Add search results to response object
if hasattr(response, "choices") and response.choices:
for choice in response.choices:

55
tests/test_litellm/integrations/vector_store_integrations/test_vector_store_pre_call_hook.py
View file

@ -11,7 +11,7 @@ from litellm.integrations.vector_store_integrations.vector_store_pre_call_hook i
ProxyServerRuntime,
VectorStorePreCallHook,
)
from litellm.types.llms.openai import AllMessageValues
from litellm.types.llms.openai import AllMessageValues, ResponsesAPIResponse
from litellm.types.utils import (
CallTypes,
Choices,
@ -363,6 +363,59 @@ async def test_a_healthy_vector_store_alone_reports_no_failures(registry_with: R
assert "vector_store_search_failures" not in (_first_message(response).provider_specific_fields or {})
@pytest.mark.asyncio
async def test_a_failing_vector_store_is_reported_on_the_responses_api_response(
registry_with: RegisterStores,
) -> None:
"""Regression (LIT-6809): /v1/responses answered 200 with no sign the knowledge base was missing."""
registry_with("vs-broken")
logging_obj = FakeLoggingObj({})
await _run_hook(
VectorStorePreCallHook(
proxy_runtime=FakeProxyRuntime(router=RecordingRouter(failing_vector_store_ids=frozenset({"vs-broken"})))
),
["vs-broken"],
logging_obj,
)
response = ResponsesAPIResponse(id="resp-lit6809", created_at=0, output=[])
await VectorStorePreCallHook(proxy_runtime=FakeProxyRuntime(router=None)).async_post_call_success_deployment_hook(
request_data={"litellm_logging_obj": logging_obj},
response=response,
call_type=CallTypes.aresponses,
)
assert response.model_dump()["vector_store_search_failures"] == [
{
"vector_store_id": "vs-broken",
"custom_llm_provider": "bedrock",
"error": "litellm.BadRequestError: no healthy deployments for vs-broken",
}
]
@pytest.mark.asyncio
async def test_a_healthy_vector_store_leaves_the_responses_api_response_alone(registry_with: RegisterStores) -> None:
registry_with("vs-healthy")
logging_obj = FakeLoggingObj({})
await _run_hook(
VectorStorePreCallHook(proxy_runtime=FakeProxyRuntime(router=RecordingRouter())),
["vs-healthy"],
logging_obj,
)
response = ResponsesAPIResponse(id="resp-lit6809", created_at=0, output=[])
await VectorStorePreCallHook(proxy_runtime=FakeProxyRuntime(router=None)).async_post_call_success_deployment_hook(
request_data={"litellm_logging_obj": logging_obj},
response=response,
call_type=CallTypes.aresponses,
)
assert "vector_store_search_failures" not in response.model_dump()
@pytest.mark.asyncio
async def test_a_failing_vector_store_is_reported_on_the_streaming_chunk(registry_with: RegisterStores) -> None:
registry_with("vs-broken")