scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-11-12 08:42:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 207

fix(embedding): reject mock input counts a provider would reject

Browse source 
An empty input list or more than 2048 inputs now raises BadRequestError
instead of returning zero rows or one row per item without limit, matching
OpenAI's documented bounds
This commit is contained in:
Aasif-Simpplr 2026-10-05 15:50:47 +05:30
parent 3d71e846b9
commit 44a2e74123
2 changed files with 25 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

27
litellm/litellm_core_utils/mock_functions.py
View file

@ -1,5 +1,7 @@
from collections.abc import Sequence
from typing import Final
from ..exceptions import BadRequestError
from ..types.utils import (
Embedding,
EmbeddingResponse,
@ -8,13 +10,15 @@ from ..types.utils import (
Usage,
)
# OpenAI caps an embedding input array at 2048 items, as of 2026-10-05:
# https://platform.openai.com/docs/api-reference/embeddings/create
_MAX_EMBEDDING_INPUTS: Final = 2048
def _embedding_input_count(embedding_input: str | Sequence[str] | Sequence[int] | Sequence[Sequence[int]]) -> int:
match embedding_input:
case str() | [int(), *_]:
return 1
case _:
return len(embedding_input)
if isinstance(embedding_input, str) or (len(embedding_input) > 0 and isinstance(embedding_input[0], int)):
return 1
return len(embedding_input)
def mock_embedding(
@ -26,12 +30,17 @@ def mock_embedding(
mock_response = [0.0] * 1536
elif mock_response == "error":
raise Exception("Mock error")
input_count: Final = _embedding_input_count(embedding_input)
if not 1 <= input_count <= _MAX_EMBEDDING_INPUTS:
raise BadRequestError(
message=f"input must contain between 1 and {_MAX_EMBEDDING_INPUTS} items, got {input_count}",
model=model,
llm_provider=None,
body={"param": "input"},
)
return EmbeddingResponse(
model=model,
data=[
Embedding(embedding=mock_response, index=index, object="embedding")
for index in range(_embedding_input_count(embedding_input))
],
data=[Embedding(embedding=mock_response, index=index, object="embedding") for index in range(input_count)],
usage=Usage(prompt_tokens=10, completion_tokens=0),
)

7
tests/unit/litellm_core_utils/test_mock_functions.py
View file

@ -3,6 +3,7 @@ from typing import Final
import pytest
import litellm
from litellm.exceptions import BadRequestError
@pytest.mark.parametrize(
@ -21,3 +22,9 @@ def test_mock_embedding_returns_one_embedding_per_input(embedding_input: object,
assert [row["index"] for row in response.data] == list(range(expected_count))
assert [row["embedding"] for row in response.data] == [vector] * expected_count
@pytest.mark.parametrize("embedding_input", [[], ["text"] * 2049])
def test_mock_embedding_rejects_input_count_a_provider_would_reject(embedding_input: list[str]) -> None:
with pytest.raises(BadRequestError, match="between 1 and 2048 items"):
litellm.embedding(model="openai/test-model", input=embedding_input, mock_response=[0.1, 0.2])