fix(vector-stores): reject searches without a query

litellm/proxy/vector_store_endpoints/endpoints.py

@@ -159,6 +159,8 @@ async def vector_store_search(
     )
 
     data = await _read_request_body(request=request)
+    if "query" not in data:
+        raise HTTPException(status_code=400, detail={"error": "query is required"})
     reject_caller_embedding_selection_params(payload=data, source="the search request body")
     data["vector_store_id"] = vector_store_id
 

tests/e2e/llm_translation/test_vector_stores_e2e.py

@@ -200,27 +200,16 @@ class TestVectorStores:
         assert deleted.id == created.id
         assert deleted.deleted is True
 
-    @pytest.mark.skip(
-        reason="stage red: product gap, vector store search 500s (asearch TypeError) on missing query instead of 400"
-    )
+    @pytest.mark.parametrize("prefix", ["/v1", ""])
     @pytest.mark.covers("llm.vector_stores.openai.input_validation.nonstream.works")
-    def test_search_missing_query_returns_error(self, proxy: ProxyClient, resources: ResourceManager) -> None:
-        key = resources.key()
-        created = unwrap(
-            proxy.transport.post(
-                "/v1/vector_stores",
-                headers=proxy.transport.bearer(key),
-                json=VectorStoreCreateBody(name=f"e2e-vs-search-{unique_marker()}"),
-                response_type=VectorStoreObject,
-            )
-        )
-        _delete_store_later(proxy, resources, key, created.id)
+    def test_search_missing_query_returns_error(self, proxy: ProxyClient, scoped_key: str, prefix: str) -> None:
         result = proxy.transport.send(
-            f"/v1/vector_stores/{created.id}/search",
-            headers=proxy.transport.bearer(key),
+            f"{prefix}/vector_stores/vs_query_validation/search",
+            headers=proxy.transport.bearer(scoped_key),
             json=VectorStoreSearchBody(max_num_results=10),
         )
-        assert_client_error(result, "vector store search missing query")
+        assert result.status_code == 400, result.body
+        assert "query is required" in result.body
 
     @pytest.mark.covers("llm.vector_stores.openai.basic.nonstream.works")
     def test_file_attach_poll_and_search(self, proxy: ProxyClient, resources: ResourceManager) -> None:

tests/test_litellm/proxy/vector_store_endpoints/test_vector_store_endpoints.py

@@ -1,5 +1,6 @@
 import json
 from datetime import datetime, timezone
+from typing import Final
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
@@ -3556,3 +3557,22 @@ def test_vector_store_search_rejects_caller_embedding_selection_params(blocked_k
 
     assert response.status_code == 400, response.json()
     assert blocked_key in str(response.json())
+
+
+@pytest.mark.parametrize("prefix", ["/v1", ""])
+def test_vector_store_search_missing_query_returns_400(prefix: str) -> None:
+    from fastapi import FastAPI
+    from fastapi.testclient import TestClient
+
+    from litellm.proxy.auth.user_api_key_auth import user_api_key_auth
+    from litellm.proxy.vector_store_endpoints.endpoints import router
+
+    app: Final = FastAPI()
+    app.include_router(router)
+    app.dependency_overrides[user_api_key_auth] = lambda: UserAPIKeyAuth(user_role=LitellmUserRoles.PROXY_ADMIN)
+    response: Final = TestClient(app, raise_server_exceptions=False).post(
+        f"{prefix}/vector_stores/documents/search", json={"max_num_results": 2}
+    )
+
+    assert response.status_code == 400, response.text
+    assert "query" in response.text