scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-13 23:11:40 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 101

refactor(router): scope cost float-cast to model-group merge comparison

Browse source
This commit is contained in:
Devin AI 2026-07-10 16:33:14 +00:00
parent 1eae186b1c
commit 783bad54ef
2 changed files with 19 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
litellm/router.py
View file

@ -8720,8 +8720,8 @@ class Router:
# Get mode from database model_info if available, otherwise default to "chat"
db_model_info = model.get("model_info", {})
mode = db_model_info.get("mode", "chat")
input_cost_per_token = safe_cast_to_float(db_model_info.get("input_cost_per_token"))
output_cost_per_token = safe_cast_to_float(db_model_info.get("output_cost_per_token"))
input_cost_per_token = db_model_info.get("input_cost_per_token")
output_cost_per_token = db_model_info.get("output_cost_per_token")
model_info = ModelMapInfo(
key=model_group,

31
tests/test_litellm/test_router.py
View file

@ -1455,11 +1455,12 @@ def test_model_group_info_cost_none_when_db_model_info_has_no_cost():
assert result.output_cost_per_token is None
def test_model_group_info_casts_string_cost_from_model_info():
def test_model_group_info_casts_string_cost_from_db_model_info():
"""
Regression for #32787: cost values round-tripped through the DB
(store_model_in_db) can come back as strings such as "1e-05". Comparing a
str against a float in _set_model_group_info raised
(store_model_in_db) can come back as strings such as "1e-05". When
get_deployment_model_info falls back to the raw db model_info, comparing a
str against a float across deployments in _set_model_group_info raised
"TypeError: '>' not supported between instances of 'str' and 'float'".
Costs read from model_info must be cast to float.
"""
@ -1470,32 +1471,34 @@ def test_model_group_info_casts_string_cost_from_model_info():
{
"model_name": "grp",
"litellm_params": {"model": "openai/a", "api_key": "fake"},
"model_info": {"id": "d1"},
"model_info": {
"id": "d1",
"input_cost_per_token": "1e-05",
"output_cost_per_token": "2e-05",
},
},
{
"model_name": "grp",
"litellm_params": {"model": "openai/b", "api_key": "fake"},
"model_info": {"id": "d2"},
"model_info": {
"id": "d2",
"input_cost_per_token": "2e-05",
"output_cost_per_token": "4e-05",
},
},
]
)
def fake_model_info(model_id, model_name):
return {
"input_cost_per_token": "1e-05",
"output_cost_per_token": "2e-05",
}
with patch.object(
router, "get_deployment_model_info", side_effect=fake_model_info
router, "get_deployment_model_info", side_effect=Exception("not found")
):
result = router._cached_get_model_group_info("grp")
assert result is not None
assert isinstance(result.input_cost_per_token, float)
assert isinstance(result.output_cost_per_token, float)
assert result.input_cost_per_token == 1e-05
assert result.output_cost_per_token == 2e-05
assert result.input_cost_per_token == 2e-05
assert result.output_cost_per_token == 4e-05
def test_get_model_access_groups_caching():