Merge 44a2e74123 into 527d882508

litellm/litellm_core_utils/mock_functions.py

@@ -1,3 +1,7 @@
+from collections.abc import Sequence
+from typing import Final
+
+from ..exceptions import BadRequestError
 from ..types.utils import (
     Embedding,
     EmbeddingResponse,
@@ -6,15 +10,37 @@ 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 mock_embedding(model: str, mock_response: list[float] | None):
+
+def _embedding_input_count(embedding_input: str | Sequence[str] | Sequence[int] | Sequence[Sequence[int]]) -> int:
+    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(
+    model: str,
+    embedding_input: str | Sequence[str] | Sequence[int] | Sequence[Sequence[int]],
+    mock_response: list[float] | None,
+):
     if mock_response is None:
         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=0, object="embedding")],
+        data=[Embedding(embedding=mock_response, index=index, object="embedding") for index in range(input_count)],
         usage=Usage(prompt_tokens=10, completion_tokens=0),
     )
 

litellm/main.py

@@ -6971,7 +6971,7 @@ def embedding(
     )
 
     if mock_response is not None:
-        return mock_embedding(model=model, mock_response=mock_response)
+        return mock_embedding(model=model, embedding_input=input, mock_response=mock_response)
     try:
         response: EmbeddingResponse | Coroutine[object, object, EmbeddingResponse] | None = None
 

tests/unit/litellm_core_utils/test_mock_functions.py

@@ -0,0 +1,30 @@
+from typing import Final
+
+import pytest
+
+import litellm
+from litellm.exceptions import BadRequestError
+
+
+@pytest.mark.parametrize(
+    ("embedding_input", "expected_count"),
+    [
+        ("hello", 1),
+        (["first", "second", "third"], 3),
+        ([101, 102, 103], 1),
+        ([[101, 102], [103, 104]], 2),
+    ],
+)
+def test_mock_embedding_returns_one_embedding_per_input(embedding_input: object, expected_count: int) -> None:
+    vector: Final = [0.1, 0.2]
+
+    response: Final = litellm.embedding(model="openai/test-model", input=embedding_input, mock_response=vector)
+
+    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])