test(vector-stores): cover S3 ingest embedding model

tests/test_litellm/proxy/rag_endpoints/test_rag_endpoints.py

@@ -6,6 +6,7 @@ Covers:
 """
 
 import io
+import json
 from unittest.mock import AsyncMock, MagicMock, patch
 
 import pytest
@@ -47,6 +48,41 @@ def client_internal_user():
         app.dependency_overrides = original_overrides
 
 
+@pytest.mark.asyncio
+async def test_s3_rag_ingest_persists_embedding_model_for_managed_search():
+    from litellm.proxy.rag_endpoints.endpoints import (
+        _save_vector_store_to_db_from_rag_ingest,
+    )
+
+    table = MagicMock()
+    table.find_unique = AsyncMock(return_value=None)
+    created_row = MagicMock()
+    created_row.model_dump.return_value = {
+        "vector_store_id": "documents:index",
+        "custom_llm_provider": "s3_vectors",
+    }
+    table.create = AsyncMock(return_value=created_row)
+    prisma_client = MagicMock()
+    prisma_client.db.litellm_managedvectorstorestable = table
+
+    await _save_vector_store_to_db_from_rag_ingest(
+        response={"vector_store_id": "documents:index"},
+        ingest_options={
+            "embedding": {"model": "text-embedding-3-large"},
+            "vector_store": {
+                "custom_llm_provider": "s3_vectors",
+                "vector_bucket_name": "documents",
+            },
+        },
+        prisma_client=prisma_client,
+        user_api_key_dict=UserAPIKeyAuth(user_id="user-1", team_id="team-1"),
+    )
+
+    persisted_params = json.loads(table.create.await_args.kwargs["data"]["litellm_params"])
+    assert persisted_params["embedding_model"] == "text-embedding-3-large"
+    assert persisted_params["vector_bucket_name"] == "documents"
+
+
 def test_internal_user_viewer_rag_ingest_without_vector_store_id_rejected(
     client_internal_user_viewer,
 ):

ui/litellm-dashboard/src/components/networking.test.ts

@@ -806,3 +806,55 @@ describe("daily activity api_key filter", () => {
     expect(requestedUrl(mockFetch)).toContain("user_id=");
   });
 });
+
+describe("ragIngestCall", () => {
+  const originalFetch = global.fetch;
+
+  afterEach(() => {
+    global.fetch = originalFetch;
+  });
+
+  const sendRequest = async (provider: string, providerParams: Record<string, unknown>) => {
+    const mockFetch = vi.fn<typeof fetch>().mockResolvedValue(
+      new Response(JSON.stringify({ status: "completed" }), {
+        status: 200,
+        headers: { "Content-Type": "application/json" },
+      }),
+    );
+    global.fetch = mockFetch;
+
+    await Networking.ragIngestCall(
+      "sk-key",
+      new File(["content"], "document.txt"),
+      provider,
+      undefined,
+      undefined,
+      undefined,
+      providerParams,
+    );
+
+    const body = mockFetch.mock.calls[0][1]?.body as FormData;
+    return JSON.parse(String(body.get("request")));
+  };
+
+  it("serializes embedding_model at the provider-defined ingest level", async () => {
+    const s3Request = await sendRequest("s3_vectors", {
+      vector_bucket_name: "documents",
+      embedding_model: "text-embedding-3-large",
+    });
+    const bedrockRequest = await sendRequest("bedrock", {
+      embedding_model: "amazon.titan-embed-text-v2:0",
+    });
+
+    expect(s3Request).toEqual({
+      ingest_options: {
+        embedding: { model: "text-embedding-3-large" },
+        vector_store: {
+          custom_llm_provider: "s3_vectors",
+          vector_bucket_name: "documents",
+        },
+      },
+    });
+    expect(bedrockRequest.ingest_options.vector_store.embedding_model).toBe("amazon.titan-embed-text-v2:0");
+  });
+});