fix(cohere): route COHERE_API_BASE through ensure_api_route and pin the url semantics

Review follow-up on #50249: COHERE_API_BASE means different things across
cohere's own paths (chat reads it as the full /v2/chat endpoint, rerank as
a base URL) and the PR initially used the verbatim reading — which breaks
a bare-host gateway silently (the request would land on the gateway root).

get_complete_url now runs whichever value it receives through
ensure_api_route(..., "v2/embed"): a bare host joins the route, a full
/v2/embed endpoint passes verbatim, and the no-env default is unchanged.
The default domain (api.cohere.ai vs the api.cohere.com the siblings use)
is left alone on purpose — flagging it here for a maintainer call instead
of bundling a behavior change into this fix.

Tests: three get_complete_url unit pins (bare host joins, full endpoint
verbatim, default unchanged) plus the env-routed respx test reshaped to
the bare-host form. 76 passed in tests/unit/llms/cohere/embed/.

Signed-off-by: 郑润泽 <506166760@qq.com>

litellm/llms/cohere/embed/transformation.py

@@ -21,6 +21,7 @@ from pydantic import BaseModel, ConfigDict, Field, TypeAdapter
 import litellm
 from litellm import COHERE_DEFAULT_EMBEDDING_INPUT_TYPE
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+from litellm.litellm_core_utils.url_utils import ensure_api_route
 from litellm.llms.base_llm import BaseEmbeddingConfig
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.types.llms.bedrock import (
@@ -205,7 +206,10 @@ class CohereEmbeddingConfig(BaseEmbeddingConfig):
         litellm_params: dict,
         stream: bool | None = None,
     ) -> str:
-        return api_base or "https://api.cohere.ai/v2/embed"
+        # ensure_api_route keeps the route-exists form working verbatim and
+        # joins the route onto a bare host, so a gateway host and a full
+        # endpoint are both usable as COHERE_API_BASE (see #50248).
+        return ensure_api_route(api_base or "https://api.cohere.ai/v2/embed", "v2/embed")
 
     def _transform_request(
         self, model: str, input: list[str], inference_params: dict

tests/unit/llms/cohere/embed/test_cohere_embedding_handler.py

@@ -34,6 +34,7 @@ async def test_bedrock_cohere_embedding_sends_the_caller_timeout(
     assert route.calls.last.request.extensions["timeout"]["read"] == 45
     assert response.data[0]["embedding"] == [0.1, 0.2]
 
+
 COHERE_DEFAULT_EMBED_URL: Final = "https://api.cohere.ai/v2/embed"
 
 
@@ -42,16 +43,14 @@ async def test_cohere_embedding_honors_cohere_api_base_env(
 ) -> None:
     """`COHERE_API_BASE` must be honored for embeddings, like the rerank path and
     the databricks/hosted_vllm embedding branches already do (#49411 family)."""
-    gateway: Final = "https://gateway.example.com/v1/embed"
-    monkeypatch.setattr(litellm, 'disable_aiohttp_transport', True)
+    gateway: Final = "https://gateway.example.com"
+    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
     monkeypatch.setenv("COHERE_API_BASE", gateway)
-    route: Final = respx_mock.post(gateway).mock(
+    route: Final = respx_mock.post(f"{gateway}/v2/embed").mock(
         return_value=httpx.Response(200, json=COHERE_EMBED_RESPONSE)
     )
 
-    response: Final = await litellm.aembedding(
-        model="cohere/embed-english-v3.0", input=["hello"], api_key="mock-key"
-    )
+    response: Final = await litellm.aembedding(model="cohere/embed-english-v3.0", input=["hello"], api_key="mock-key")
 
     assert route.calls.call_count == 1
     assert response.data[0]["embedding"] == [0.1, 0.2]
@@ -60,15 +59,13 @@ async def test_cohere_embedding_honors_cohere_api_base_env(
 async def test_cohere_embedding_keeps_the_default_url_without_the_env(
     respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
 ) -> None:
-    monkeypatch.setattr(litellm, 'disable_aiohttp_transport', True)
+    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
     monkeypatch.delenv("COHERE_API_BASE", raising=False)
     route: Final = respx_mock.post(COHERE_DEFAULT_EMBED_URL).mock(
         return_value=httpx.Response(200, json=COHERE_EMBED_RESPONSE)
     )
 
-    response: Final = await litellm.aembedding(
-        model="cohere/embed-english-v3.0", input=["hello"], api_key="mock-key"
-    )
+    response: Final = await litellm.aembedding(model="cohere/embed-english-v3.0", input=["hello"], api_key="mock-key")
 
     assert route.calls.call_count == 1
     assert response.data[0]["embedding"] == [0.1, 0.2]

tests/unit/llms/cohere/embed/test_transformation.py

@@ -443,3 +443,41 @@ async def test_cohere_embed_v4_image_embedding(respx_mock: respx.MockRouter, syn
         "input_type": "image",
     }
     assert response.data == [{"object": "embedding", "index": 0, "embedding": [0.125, -0.375]}]
+
+
+# --------------------------------------------------------------------------------------
+# get_complete_url: COHERE_API_BASE routing semantics (#50248 follow-up)
+# --------------------------------------------------------------------------------------
+
+
+def test_get_complete_url_appends_the_route_to_a_bare_host() -> None:
+    url: Final = CohereEmbeddingConfig().get_complete_url(
+        api_base="https://gateway.example.com",
+        api_key=None,
+        model="embed-english-v3.0",
+        optional_params={},
+        litellm_params={},
+    )
+    assert url == "https://gateway.example.com/v2/embed"
+
+
+def test_get_complete_url_keeps_a_full_endpoint_verbatim() -> None:
+    url: Final = CohereEmbeddingConfig().get_complete_url(
+        api_base="https://gateway.example.com/v2/embed",
+        api_key=None,
+        model="embed-english-v3.0",
+        optional_params={},
+        litellm_params={},
+    )
+    assert url == "https://gateway.example.com/v2/embed"
+
+
+def test_get_complete_url_default_is_unchanged() -> None:
+    url: Final = CohereEmbeddingConfig().get_complete_url(
+        api_base=None,
+        api_key=None,
+        model="embed-english-v3.0",
+        optional_params={},
+        litellm_params={},
+    )
+    assert url == "https://api.cohere.ai/v2/embed"