Add list live model names endpoint

tensorflow_serving/apis/get_model_status.proto

@@ -16,6 +16,16 @@ message GetModelStatusRequest {
   ModelSpec model_spec = 1;
 }
 
+//request to get all live model names
+message ListModelNamesRequest{
+}
+
+//response contains all live model names
+message ListModelNamesResponse{
+  repeated string model_names = 1;
+}
+
+
 // Version number, state, and status for a single version of a model.
 message ModelVersionStatus {
   // Model version.

tensorflow_serving/apis/model_service.proto

@@ -21,4 +21,7 @@ service ModelService {
   // longer served.
   rpc HandleReloadConfigRequest(ReloadConfigRequest)
       returns (ReloadConfigResponse);
+
+  // Lists the live model names
+  rpc ListModelNames(ListModelNamesRequest) returns (ListModelNamesResponse);
 }

tensorflow_serving/model_servers/get_model_status_impl.cc

@@ -62,6 +62,12 @@ void AddModelVersionStatusToResponse(GetModelStatusResponse* response,
 
 }  // namespace
 
+// Add model name to ListModelNamesResponse
+void AddModelNameToResponse(ListModelNamesResponse* response,
+			    const string& name){
+  response->add_model_names(name);
+}
+
 Status GetModelStatusImpl::GetModelStatus(ServerCore* core,
                                           const GetModelStatusRequest& request,
                                           GetModelStatusResponse* response) {
@@ -106,5 +112,16 @@ Status GetModelStatusImpl::GetModelStatusWithModelSpec(
   return tensorflow::Status::OK();
 }
 
+Status GetModelStatusImpl::ListModelNames(ServerCore* core,
+                                          const ListModelNamesRequest& request,
+                                          ListModelNamesResponse* response) {
+
+  const ServableStateMonitor& monitor = *core->servable_state_monitor();
+  for(const auto&  servable: monitor.GetAllServableStates()) {
+	  AddModelNameToResponse(response, servable.first);
+  }
+  return tensorflow::Status::OK();
+}
+
 }  // namespace serving
 }  // namespace tensorflow

tensorflow_serving/model_servers/get_model_status_impl.h

@@ -38,8 +38,12 @@ class GetModelStatusImpl {
   static Status GetModelStatusWithModelSpec(
       ServerCore* core, const ModelSpec& model_spec,
       const GetModelStatusRequest& request, GetModelStatusResponse* response);
-};
 
+  // List all live model names
+  static Status ListModelNames(ServerCore* core,
+                               const ListModelNamesRequest& request,
+                               ListModelNamesResponse* response);
+};
 }  // namespace serving
 }  // namespace tensorflow
 

tensorflow_serving/model_servers/get_model_status_impl_test.cc

@@ -172,6 +172,17 @@ TEST_F(GetModelStatusImplTest, SingleVersionSuccess) {
   EXPECT_EQ("", response.model_version_status(0).status().error_message());
 }
 
+TEST_F(GetModelStatusImplTest, Success) {
+  ListModelsRequest request;
+  ListModelsResponse response;
+  // If two versions of model are managed by ServerCore, succesfully get model
+  // status for both versions of the model.
+  TF_EXPECT_OK(
+      GetModelStatusImpl::ListModels(GetServerCore(), request, &response)
+      );
+  EXPECT_EQ(1, response.model_names_size());
+}
+
 // Verifies that GetModelStatusWithModelSpec() uses the model spec override
 // rather than the one in the request.
 TEST_F(GetModelStatusImplTest, ModelSpecOverride) {