fix(vector-store): build the Router executor from the router kwarg and gate the SDK fallback on explicit config

litellm/llms/base_llm/vector_store/transformation.py

@@ -86,9 +86,12 @@ class RouterVectorStoreEmbeddingExecutor:
         )
         return bool(resolved) or model in deployment_models
 
+    def _embeds_through_sdk(self, model: str, configuration: Mapping[str, object]) -> bool:
+        return bool(configuration) and not self._router_serves(model)
+
     def embed(self, model: str, query: str, configuration: Mapping[str, object]) -> EmbeddingResponse:
         embedding_kwargs: Final = self._embedding_kwargs(configuration)
-        if not self._router_serves(model):
+        if self._embeds_through_sdk(model, configuration):
             return LiteLLMVectorStoreEmbeddingExecutor().embed(model, query, embedding_kwargs)
         return self.router.embedding(  # pyright: ignore[reportUnknownMemberType]  # Router embedding input retains a legacy untyped list
             model=model,
@@ -98,7 +101,7 @@ class RouterVectorStoreEmbeddingExecutor:
 
     async def aembed(self, model: str, query: str, configuration: Mapping[str, object]) -> EmbeddingResponse:
         embedding_kwargs: Final = self._embedding_kwargs(configuration)
-        if not self._router_serves(model):
+        if self._embeds_through_sdk(model, configuration):
             return await LiteLLMVectorStoreEmbeddingExecutor().aembed(model, query, embedding_kwargs)
         return await self.router.aembedding(  # pyright: ignore[reportUnknownMemberType]  # Router embedding input retains a legacy untyped list
             model=model,

litellm/vector_stores/main.py

@@ -16,7 +16,7 @@ from litellm.constants import request_timeout
 from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.llms.base_llm.vector_store.transformation import (
-    LiteLLMVectorStoreEmbeddingExecutor,
+    BaseQueryEmbeddingVectorStoreConfig,
     VectorStoreEmbeddingExecutor,
 )
 from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler
@@ -42,12 +42,10 @@ base_llm_http_handler = BaseLLMHTTPHandler()
 #################################################
 
 
-def _direct_vector_store_embedding_executor(value: object) -> VectorStoreEmbeddingExecutor:
-    if value is None:
-        return LiteLLMVectorStoreEmbeddingExecutor()
-    if isinstance(value, VectorStoreEmbeddingExecutor):
-        return value
-    raise TypeError("Invalid direct vector store embedding executor")
+def _direct_vector_store_embedding_executor(value: object, router: "Router | None") -> VectorStoreEmbeddingExecutor:
+    if value is not None and not isinstance(value, VectorStoreEmbeddingExecutor):
+        raise TypeError("Invalid direct vector store embedding executor")
+    return BaseQueryEmbeddingVectorStoreConfig.query_embedding_executor(value, router)
 
 
 def mock_vector_store_search_response(
@@ -302,7 +300,7 @@ async def asearch(
     Async: Search a vector store for relevant chunks based on a query and file attributes filter.
     """
     embedding_executor: Final = _direct_vector_store_embedding_executor(
-        kwargs.pop("_direct_vector_store_embedding_executor", None)
+        kwargs.pop("_direct_vector_store_embedding_executor", None), router
     )
     local_vars: Final = {  # mutable-ok: exception logging requires a sanitized mutable snapshot
         key: value for key, value in locals().items() if key != "embedding_executor"
@@ -388,7 +386,7 @@ def search(
         VectorStoreSearchResponse containing the search results.
     """
     embedding_executor: Final = _direct_vector_store_embedding_executor(
-        kwargs.pop("_direct_vector_store_embedding_executor", None)
+        kwargs.pop("_direct_vector_store_embedding_executor", None), router
     )
     local_vars: Final = {  # mutable-ok: exception logging requires a sanitized mutable snapshot
         key: value for key, value in locals().items() if key != "embedding_executor"

tests/router_unit_tests/test_router_embedding_integration.py

@@ -186,6 +186,25 @@ class TestRouterEmbeddingIntegration:
         assert _sent(store_route, 0) == ("Bearer store-key", "text-embedding-3-large", ["sync query"])
         assert _sent(store_route, 1) == ("Bearer store-key", "text-embedding-3-large", ["async query"])
 
+    @pytest.mark.asyncio
+    async def test_router_executor_rejects_unserved_models_without_explicit_config(
+        self, respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
+    ):
+        monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
+        monkeypatch.setenv("OPENAI_API_KEY", "env-key")
+        openai_route = _mock_embedding_route(respx_mock, OPENAI_EMBEDDINGS_URL)
+        executor = RouterVectorStoreEmbeddingExecutor(
+            router=_alias_router(),
+            metadata={"user_api_key_team_id": "team-a"},
+        )
+
+        with pytest.raises(litellm.BadRequestError):
+            executor.embed("openai/text-embedding-3-large", "sync query", {})
+        with pytest.raises(litellm.BadRequestError):
+            await executor.aembed("openai/text-embedding-3-large", "async query", {})
+
+        assert openai_route.call_count == 0
+
     def test_router_executor_routes_deployment_model_names_through_the_router(
         self, respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
     ):

tests/test_litellm/proxy/vector_store_endpoints/test_vector_store_endpoints.py

@@ -52,7 +52,7 @@ def _serialize_litellm_params(litellm_params):
 
 def test_direct_vector_store_embedding_executor_rejects_invalid_value():
     with pytest.raises(TypeError, match="Invalid direct vector store embedding executor"):
-        _direct_vector_store_embedding_executor(object())
+        _direct_vector_store_embedding_executor(object(), None)
 
 
 def test_router_vector_store_search_injects_executor_and_request_metadata():

tests/vector_store_tests/test_milvus_vector_store.py

@@ -575,6 +575,32 @@ async def test_router_search_resolves_bare_embedding_alias_async(
     _assert_alias_resolved(embedding_route, search_route, response)
 
 
+def test_sdk_search_with_router_kwarg_resolves_bare_embedding_alias_sync(
+    respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
+):
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    embedding_route = _mock_embedding_route(respx_mock)
+    search_route = _mock_search_route(respx_mock)
+
+    response = litellm.vector_stores.search(router=_alias_router(), **ALIAS_SEARCH_KWARGS)
+
+    _assert_alias_resolved(embedding_route, search_route, response)
+
+
+@pytest.mark.asyncio
+async def test_sdk_search_with_router_kwarg_resolves_bare_embedding_alias_async(
+    respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
+):
+    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    embedding_route = _mock_embedding_route(respx_mock)
+    search_route = _mock_search_route(respx_mock)
+
+    response = await litellm.vector_stores.asearch(router=_alias_router(), **ALIAS_SEARCH_KWARGS)
+
+    _assert_alias_resolved(embedding_route, search_route, response)
+
+
 @pytest.mark.asyncio
 async def test_transform_uses_injected_executor_without_embedding_config(respx_mock: respx.MockRouter):
     executor = RecordingEmbeddingExecutor(ALIAS_EMBEDDING_RESPONSE)