fix: preserve Google embedding usage metadata

lib/req_llm/providers/google.ex

@@ -1134,12 +1134,14 @@ defmodule ReqLLM.Providers.Google do
     raise ReqLLM.Error.Invalid.Parameter, parameter: "schema: #{message}"
   end
 
-  defp normalize_embedding_response(%{"embedding" => %{"values" => values}})
+  defp normalize_embedding_response(%{"embedding" => %{"values" => values}} = body)
        when is_list(values) do
     %{"data" => [%{"index" => 0, "embedding" => values}]}
+    |> maybe_put_embedding_usage_metadata(body)
   end
 
-  defp normalize_embedding_response(%{"embeddings" => embeddings}) when is_list(embeddings) do
+  defp normalize_embedding_response(%{"embeddings" => embeddings} = body)
+       when is_list(embeddings) do
     data =
       embeddings
       |> Enum.with_index()
@@ -1153,10 +1155,21 @@ defmodule ReqLLM.Providers.Google do
       end)
 
     %{"data" => data}
+    |> maybe_put_embedding_usage_metadata(body)
   end
 
   defp normalize_embedding_response(other), do: other
 
+  defp maybe_put_embedding_usage_metadata(normalized, body) do
+    case Map.get(body, "usageMetadata") do
+      usage_metadata when is_map(usage_metadata) ->
+        Map.put(normalized, "usageMetadata", usage_metadata)
+
+      _ ->
+        normalized
+    end
+  end
+
   @impl ReqLLM.Provider
   def decode_response({req, resp}) do
     case resp.status do

test/providers/google_test.exs

@@ -747,6 +747,10 @@ defmodule ReqLLM.Providers.GoogleTest do
       embedding_response = %{
         "embedding" => %{
           "values" => [0.1, -0.2, 0.3, 0.4, -0.5]
+        },
+        "usageMetadata" => %{
+          "promptTokenCount" => 2,
+          "totalTokenCount" => 2
         }
       }
 
@@ -764,7 +768,11 @@ defmodule ReqLLM.Providers.GoogleTest do
       assert req == mock_req
 
       assert resp.body == %{
-               "data" => [%{"index" => 0, "embedding" => [0.1, -0.2, 0.3, 0.4, -0.5]}]
+               "data" => [%{"index" => 0, "embedding" => [0.1, -0.2, 0.3, 0.4, -0.5]}],
+               "usageMetadata" => %{
+                 "promptTokenCount" => 2,
+                 "totalTokenCount" => 2
+               }
              }
     end
 

test/req_llm/embedding_test.exs

@@ -181,6 +181,48 @@ defmodule ReqLLM.EmbeddingTest do
     end
   end
 
+  describe "embed/3 - Google usage metadata" do
+    setup do
+      Req.Test.stub(__MODULE__.GoogleEmbedUsage, fn conn ->
+        Req.Test.json(conn, %{
+          "embedding" => %{
+            "values" => [0.1, -0.2, 0.3]
+          },
+          "usageMetadata" => %{
+            "promptTokenCount" => 2,
+            "totalTokenCount" => 2
+          }
+        })
+      end)
+
+      setup_telemetry()
+    end
+
+    test "emits telemetry and returns usage for embeddings" do
+      {:ok, %{embedding: embedding, usage: usage}} =
+        Embedding.embed(
+          "google:gemini-embedding-001",
+          "Hello world",
+          api_key: "test-key",
+          return_usage: true,
+          req_http_options: [plug: {Req.Test, __MODULE__.GoogleEmbedUsage}]
+        )
+
+      assert embedding == [0.1, -0.2, 0.3]
+      assert usage.input == 2
+      assert usage.total_tokens == 2
+
+      assert_receive {:telemetry_event, [:req_llm, :token_usage], measurements,
+                      %{model: %LLMDB.Model{provider: :google, id: "gemini-embedding-001"}} =
+                        metadata}
+
+      assert measurements.tokens.input == 2
+      assert measurements.tokens.total_tokens == 2
+      assert metadata.model.provider == :google
+      assert metadata.model.id == "gemini-embedding-001"
+    end
+  end
+
   describe "embed_many/3 - basic functionality" do
     test "validates model before attempting embedding" do
       case Embedding.validate_model("openai:text-embedding-3-small") do