fix(main.py): forward headers to multimodal embeddings and refresh lock file

litellm/main.py

@@ -5148,6 +5148,7 @@ def embedding(  # noqa: PLR0915
                     vertex_project=vertex_ai_project,
                     vertex_location=vertex_ai_location,
                     vertex_credentials=vertex_credentials,
+                    headers=headers,
                     aembedding=aembedding,
                     print_verbose=print_verbose,
                     custom_llm_provider="vertex_ai",

tests/test_litellm/test_vertex_ai_embedding_headers.py

@@ -3,16 +3,9 @@ from unittest.mock import patch, MagicMock
 from litellm import embedding
 
 def test_vertex_ai_embedding_extra_headers():
-    """
-    Test that extra_headers are correctly forwarded to the 
-    vertex_embedding.embedding function.
-    """
-    # We patch the exact location where main.py calls the vertex provider
-    with patch("litellm.main.vertex_embedding.embedding") as mock_vertex_embedding:
-        # Mock a successful return so the call doesn't fail
-        mock_vertex_embedding.return_value = MagicMock()
-        
-        # Trigger the embedding call
+    """Test standard vertex embedding header forwarding."""
+    with patch("litellm.main.vertex_embedding.embedding") as mock_vertex:
+        mock_vertex.return_value = MagicMock()
         try:
             embedding(
                 model="vertex_ai/text-embedding-004",
@@ -20,13 +13,24 @@ def test_vertex_ai_embedding_extra_headers():
                 extra_headers={"X-Custom-Header": "test-value"},
             )
         except Exception:
-            # We don't care about subsequent errors, only the forwarding
             pass
-        
-        # VERIFICATION: This is the important part
-        mock_vertex_embedding.assert_called_once()
-        call_kwargs = mock_vertex_embedding.call_args.kwargs
-        
-        # Check that the headers we passed actually reached the provider
-        assert call_kwargs.get("extra_headers") == {"X-Custom-Header": "test-value"}
-        print("\n✅ Success: extra_headers correctly forwarded to Vertex AI provider!")
+        mock_vertex.assert_called_once()
+        assert mock_vertex.call_args.kwargs.get("extra_headers") == {"X-Custom-Header": "test-value"}
+
+def test_vertex_multimodal_embedding_headers():
+    """Test multimodal vertex embedding header forwarding."""
+    with patch("litellm.main.vertex_multimodal_embedding.multimodal_embedding") as mock_multimodal:
+        mock_multimodal.return_value = MagicMock()
+        try:
+            # Using a multimodal model triggers the different provider path
+            embedding(
+                model="vertex_ai/multimodalembedding@001",
+                input=["hello"],
+                extra_headers={"X-Custom-Header": "multi-value"},
+            )
+        except Exception:
+            pass
+        mock_multimodal.assert_called_once()
+        # NOTE: The multimodal handler uses 'headers' as the parameter name
+        assert mock_multimodal.call_args.kwargs.get("headers") == {"X-Custom-Header": "multi-value"}
+        print("\n✅ Success: Headers forwarded for both standard and multimodal Vertex AI!")