From c254605e924455062e72305fdfaa2bd9c9feb089 Mon Sep 17 00:00:00 2001 From: mateo-berri <277851410+mateo-berri@users.noreply.github.com> Date: Sat, 29 Aug 2026 11:10:21 -0700 Subject: [PATCH] test(embeddings): move encoding_format default coverage to wire-level assertions Consolidate the new regression tests into test_openai_embedding_encoding_format_default.py, replacing mocks that pinned the old float default with respx captures of the request body, and update the stale local_testing default-float test to assert omission --- tests/local_testing/test_embedding.py | 64 +++--- tests/test_litellm/test_main.py | 60 ------ ...penai_embedding_encoding_format_default.py | 188 ++++++++---------- 3 files changed, 109 insertions(+), 203 deletions(-) diff --git a/tests/local_testing/test_embedding.py b/tests/local_testing/test_embedding.py index aed2849f056..ee2ac14f498 100644 --- a/tests/local_testing/test_embedding.py +++ b/tests/local_testing/test_embedding.py @@ -3,6 +3,8 @@ import os import re import traceback +import httpx + import openai import pytest from dotenv import load_dotenv @@ -1255,56 +1257,42 @@ def test_jina_ai_img_embeddings(input_data, expected_payload_input): assert sent_data["input"] == expected_payload_input -def test_encoding_format_defaults_to_float_for_openai_sdk(monkeypatch): +def test_encoding_format_omitted_by_default_for_openai_sdk(monkeypatch): """ - When encoding_format is not provided, LiteLLM sends `float` for OpenAI-path embeddings. + When encoding_format is not provided, LiteLLM leaves it out of the upstream request. Optional global override: `LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT`. """ monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - # Create a mock client instance - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance + captured_bodies = [] - # Mock the embeddings.with_raw_response.create method - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", + def handler(request: httpx.Request) -> httpx.Response: + captured_bodies.append(json.loads(request.content)) + return httpx.Response( + 200, + json={ "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "text-embedding-ada-002", "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } - ) - mock_response.headers = {} - - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response + }, ) - # Call the embedding function without encoding_format - response = embedding( - model="text-embedding-ada-002", - input="Hello world", - ) + client = openai.OpenAI( + api_key="sk-test", http_client=httpx.Client(transport=httpx.MockTransport(handler)) + ) - # Get the call arguments to verify what was sent to OpenAI SDK - call_args = mock_client_instance.embeddings.with_raw_response.create.call_args - assert ( - call_args is not None - ), "OpenAI SDK embeddings.create should have been called" + response = embedding( + model="text-embedding-ada-002", + input="Hello world", + api_key="sk-test", + client=client, + ) - call_kwargs = call_args[1] # Get kwargs - - assert "encoding_format" in call_kwargs - assert ( - call_kwargs["encoding_format"] == "float" - ), "encoding_format should default to float when not provided by user" - - print("✅ PASS: encoding_format='float' is correctly passed to OpenAI SDK") + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] + assert "encoding_format" not in captured_bodies[0], ( + "encoding_format should be omitted from the upstream request when not provided by user" + ) def test_encoding_format_explicit_value_preserved(): diff --git a/tests/test_litellm/test_main.py b/tests/test_litellm/test_main.py index 4c591be0f61..84225c3feb1 100644 --- a/tests/test_litellm/test_main.py +++ b/tests/test_litellm/test_main.py @@ -3182,63 +3182,3 @@ def test_stream_chunk_builder_defers_cost_to_logging_obj_when_usage_cost_absent( assert response is not None assert response._hidden_params.get("response_cost") is None - -def _mock_openai_embedding_route(respx_mock: respx.MockRouter) -> respx.Route: - return respx_mock.post("https://api.openai.com/v1/embeddings").mock( - return_value=httpx.Response( - 200, - json={ - "object": "list", - "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], - "model": "text-embedding-3-small", - "usage": {"prompt_tokens": 2, "total_tokens": 2}, - }, - ) - ) - - -def test_embedding_openai_omits_encoding_format_when_client_omits_it(respx_mock: respx.MockRouter) -> None: - mock_route: Final = _mock_openai_embedding_route(respx_mock) - - response: Final = litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") - - request_body: Final = json.loads(mock_route.calls.last.request.read()) - assert "encoding_format" not in request_body - assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] - - -def test_embedding_openai_forwards_explicit_encoding_format(respx_mock: respx.MockRouter) -> None: - mock_route: Final = _mock_openai_embedding_route(respx_mock) - - litellm.embedding( - model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", encoding_format="base64" - ) - - request_body: Final = json.loads(mock_route.calls.last.request.read()) - assert request_body["encoding_format"] == "base64" - - -def test_embedding_openai_env_var_sets_default_encoding_format( - respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch -) -> None: - monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", "float") - mock_route: Final = _mock_openai_embedding_route(respx_mock) - - litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") - - request_body: Final = json.loads(mock_route.calls.last.request.read()) - assert request_body["encoding_format"] == "float" - - -@pytest.mark.asyncio -async def test_aembedding_openai_omits_encoding_format_when_client_omits_it( - respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch -) -> None: - monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) - mock_route: Final = _mock_openai_embedding_route(respx_mock) - - response: Final = await litellm.aembedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") - - request_body: Final = json.loads(mock_route.calls.last.request.read()) - assert "encoding_format" not in request_body - assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] diff --git a/tests/test_litellm/test_openai_embedding_encoding_format_default.py b/tests/test_litellm/test_openai_embedding_encoding_format_default.py index 94e4e3c81e5..9842bf30585 100644 --- a/tests/test_litellm/test_openai_embedding_encoding_format_default.py +++ b/tests/test_litellm/test_openai_embedding_encoding_format_default.py @@ -1,124 +1,102 @@ -from unittest.mock import MagicMock, patch +import json +from typing import Final +import httpx import pytest +import respx -from litellm import embedding +import litellm -@pytest.mark.parametrize( - "set_env, env_value, expected", - [ - (False, None, "float"), - (True, "base64", "base64"), - ], -) -def test_openai_embedding_encoding_format_default( - monkeypatch, set_env, env_value, expected -): - monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - if set_env: - monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_value) - - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } +def _mock_openai_embedding_route(respx_mock: respx.MockRouter) -> respx.Route: + return respx_mock.post("https://api.openai.com/v1/embeddings").mock( + return_value=httpx.Response( + 200, + json={ + "object": "list", + "data": [{"object": "embedding", "index": 0, "embedding": [0.1, 0.2, 0.3]}], + "model": "text-embedding-3-small", + "usage": {"prompt_tokens": 2, "total_tokens": 2}, + }, + ) ) - mock_response.headers = {} - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response - ) - embedding( - model="text-embedding-ada-002", - input="Hello world", - ) +@pytest.fixture(autouse=True) +def clear_default_encoding_format_env(monkeypatch: pytest.MonkeyPatch): + monkeypatch.delenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", raising=False) - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert call_kwargs["encoding_format"] == expected + +def test_embedding_openai_omits_encoding_format_when_client_omits_it(respx_mock: respx.MockRouter) -> None: + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + response: Final = litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3] + + +def test_embedding_openai_forwards_explicit_encoding_format(respx_mock: respx.MockRouter) -> None: + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding( + model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", encoding_format="base64" + ) + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == "base64" + + +def test_embedding_openai_explicit_encoding_format_wins_over_env_var( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", "float") + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding( + model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test", encoding_format="base64" + ) + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == "base64" + + +@pytest.mark.parametrize("env_value", ["float", "base64"]) +def test_embedding_openai_env_var_sets_default_encoding_format( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch, env_value: str +) -> None: + monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_value) + mock_route: Final = _mock_openai_embedding_route(respx_mock) + + litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") + + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert request_body["encoding_format"] == env_value @pytest.mark.parametrize("env_none", ["none", "NONE", " none "]) -def test_openai_embedding_encoding_format_env_none_omits_param( - monkeypatch, env_none -): - """LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT=none omits encoding_format (provider default).""" +def test_embedding_openai_env_none_omits_encoding_format( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch, env_none: str +) -> None: monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", env_none) + mock_route: Final = _mock_openai_embedding_route(respx_mock) - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } - ) - mock_response.headers = {} + litellm.embedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response - ) - - embedding( - model="text-embedding-ada-002", - input="Hello world", - ) - - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert "encoding_format" not in call_kwargs + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body -def test_openai_embedding_encoding_format_explicit_overrides_env(monkeypatch): - """Request `encoding_format` wins over LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT.""" - monkeypatch.setenv("LITELLM_DEFAULT_EMBEDDING_ENCODING_FORMAT", "float") +@pytest.mark.asyncio +async def test_aembedding_openai_omits_encoding_format_when_client_omits_it( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + mock_route: Final = _mock_openai_embedding_route(respx_mock) - mock_response = MagicMock() - mock_response.parse.return_value = MagicMock( - model_dump=lambda: { - "data": [{"embedding": [0.1, 0.2, 0.3], "index": 0}], - "model": "text-embedding-ada-002", - "object": "list", - "usage": {"prompt_tokens": 1, "total_tokens": 1}, - } - ) - mock_response.headers = {} + response: Final = await litellm.aembedding(model="openai/text-embedding-3-small", input=["hello"], api_key="sk-test") - with patch( - "litellm.llms.openai.openai.OpenAIChatCompletion._get_openai_client" - ) as mock_get_client: - mock_client_instance = MagicMock() - mock_get_client.return_value = mock_client_instance - mock_client_instance.embeddings.with_raw_response.create.return_value = ( - mock_response - ) - - embedding( - model="text-embedding-ada-002", - input="Hello world", - encoding_format="base64", - ) - - call_kwargs = ( - mock_client_instance.embeddings.with_raw_response.create.call_args[1] - ) - assert call_kwargs["encoding_format"] == "base64" + request_body: Final = json.loads(mock_route.calls.last.request.read()) + assert "encoding_format" not in request_body + assert response.data[0]["embedding"] == [0.1, 0.2, 0.3]