scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-08 03:08:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 114

fix(proxy): authorize health test-connection on final probe params

Browse source 
Move _update_litellm_params_for_health_check before can_user_make_model_call
so health_check_params cannot retarget the probe after the auth check. Type
the Pegasus test helper signature and drop the redundant test narrative.
This commit is contained in:
mateo-berri 2026-08-24 11:24:21 -07:00
parent d0dd24ed6d
commit bf4069a80c
2 changed files with 9 additions and 17 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
litellm/proxy/health_endpoints/_health_endpoints.py
View file

@ -1952,8 +1952,13 @@ async def test_model_connection(
**request_litellm_params,
}
## Auth check
resolved_model_info: Final = loaded_model_info if loaded_model_info is not None else model_info
litellm_params = _update_litellm_params_for_health_check(
model_info=resolved_model_info or {},
litellm_params=litellm_params,
)
## Auth check, on the final probe params so health_check_params cannot retarget it afterwards
await ModelManagementAuthChecks.can_user_make_model_call(
model_params=Deployment(
model_name="test_model",
@ -1964,10 +1969,6 @@ async def test_model_connection(
prisma_client=prisma_client,
premium_user=premium_user,
)
litellm_params = _update_litellm_params_for_health_check(
model_info=resolved_model_info or {},
litellm_params=litellm_params,
)
mode = mode or litellm_params.pop("mode", None)
result: Final = await run_with_timeout(

15
tests/test_litellm/proxy/test_health_check_max_tokens.py
View file

@ -549,17 +549,6 @@ async def test_run_model_health_check_skips_auto_router_deployment():
assert result == {}
# ---------------------------------------------------------------------------
# model_info.health_check_params
#
# Some providers require a payload field litellm does not synthesize for a
# probe. Bedrock TwelveLabs Pegasus rejects any Invoke body without a top-level
# `mediaSource`, so every health check on such a deployment failed with
# "Invalid JSON: $: required property 'mediaSource' not found". The config key
# was accepted and then never read, so operators had no way to supply it.
# ---------------------------------------------------------------------------
def test_health_check_params_merge_into_probe_params():
"""health_check_params reach the probe request for the deployment that declares them."""
media_source = {"s3Location": {"uri": "s3://my-bucket/clip.mp4"}}
@ -640,7 +629,9 @@ def test_health_check_params_apply_to_non_chat_modes():
assert "max_tokens" not in updated
async def _pegasus_health_check_request_body(model_info: dict, monkeypatch) -> dict:
async def _pegasus_health_check_request_body(
model_info: dict[str, object], monkeypatch: pytest.MonkeyPatch
) -> dict[str, object]:
monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
litellm.in_memory_llm_clients_cache.flush_cache()