scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-11-06 07:48:06 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 320

fix(dashscope): keep an explicit compatible-api rerank base as given

Browse source
This commit is contained in:
mateo-berri 2026-10-28 12:32:35 -07:00
parent be9f9afecb
commit 5e83b0a363
2 changed files with 42 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
litellm/llms/dashscope/common_utils.py
View file

@ -96,12 +96,19 @@ def resolve_dashscope_endpoints(api_base: str | None) -> DashScopeEndpoints | No
)
def _is_compatible_api_base(api_base: str) -> bool:
path: Final = urlparse(api_base).path.rstrip("/")
return path == DASHSCOPE_COMPATIBLE_API_PATH or path.startswith(f"{DASHSCOPE_COMPATIBLE_API_PATH}/")
def resolve_dashscope_openai_api_base(
api_base: str | None,
env_var: str = "DASHSCOPE_API_BASE",
default: str = f"{DASHSCOPE_DEFAULT_ORIGIN}{DASHSCOPE_CHAT_COMPATIBLE_PATH}",
) -> str:
candidate: Final = api_base or get_secret_str(env_var)
if candidate and _is_compatible_api_base(candidate):
return candidate
endpoints: Final = resolve_dashscope_endpoints(candidate)
if endpoints is not None:
return endpoints.openai
@ -146,12 +153,8 @@ def _rerank_origin(api_base: str | None) -> str | None:
return endpoints.origin if endpoints is not None else _chat_shaped_aliyun_origin(api_base)
def _is_rerank_url(api_base: str) -> bool:
return urlparse(api_base).path.rstrip("/") == DASHSCOPE_RERANK_PATH
def resolve_dashscope_family_rerank_api_base(api_base: str | None, env_var: str, default_rerank_base: str) -> str:
if api_base is not None and _is_rerank_url(api_base):
if api_base is not None and _is_compatible_api_base(api_base):
return api_base
origin: Final = _rerank_origin(api_base)
if origin is not None:

34
tests/unit/llms/dashscope/test_dashscope_rerank_transformation.py
View file

@ -8,8 +8,10 @@ from unittest.mock import MagicMock
import httpx
import pytest
import respx
from pydantic import ValidationError
import litellm
from litellm.llms.dashscope.common_utils import DashScopeError
from litellm.llms.dashscope.rerank.transformation import (
@ -450,3 +452,35 @@ def test_transform_rerank_response_shape_errors_do_not_echo_the_payload():
_transform({"results": [["leaked document text"]]})
assert "leaked document text" not in str(exc_info.value)
@pytest.mark.parametrize(
("provider", "rerank_env"),
[
("dashscope", "DASHSCOPE_API_BASE_RERANK"),
("qwencloud", "QWENCLOUD_API_BASE_RERANK"),
("qwen_ai_platform", "QWEN_AI_PLATFORM_API_BASE_RERANK"),
],
)
@pytest.mark.parametrize(
"api_base",
[
"https://dashscope-intl.aliyuncs.com/compatible-api/v1/reranks",
"https://dashscope-intl.aliyuncs.com/compatible-api/v1",
],
)
def test_rerank_sends_an_explicit_compatible_api_base_over_the_rerank_env(
monkeypatch: pytest.MonkeyPatch, provider: str, rerank_env: str, api_base: str
):
monkeypatch.setenv("DASHSCOPE_API_BASE", "https://dashscope-intl.aliyuncs.com")
monkeypatch.setenv(rerank_env, DEFAULT_RERANK_URL)
with respx.mock(assert_all_called=True) as upstream:
upstream.post("https://dashscope-intl.aliyuncs.com/compatible-api/v1/reranks").respond(
200, json={"id": "r1", "results": [{"index": 0, "relevance_score": 0.9}], "usage": {"total_tokens": 5}}
)
response: Final = litellm.rerank(
model=f"{provider}/qwen3-rerank", query="q", documents=["a"], api_base=api_base, api_key="sk-test"
)
assert response.results == [{"index": 0, "relevance_score": 0.9}]