Split out worker related things to a separate service definition.

Clients/Cpp/sample.cpp

@@ -1,25 +1,11 @@
 #include "host.grpc.pb.h"
+#include "model.grpc.pb.h"
 #include "image_generation.grpc.pb.h"
 
 #include <grpc++/grpc++.h>
 
 using namespace gay::pizza::stable::diffusion;
 
-int CompareModelInfoByLoadedFirst(ModelInfo& left, ModelInfo& right) {
-    if (left.is_loaded() && right.is_loaded()) {
-        return 0;
-    }
-
-    if (left.is_loaded()) {
-        return 1;
-    }
-
-    if (right.is_loaded()) {
-        return -1;
-    }
-    return 0;
-}
-
 int main() {
     auto channel = grpc::CreateChannel("localhost:4546", grpc::InsecureChannelCredentials());
     auto modelService = ModelService::NewStub(channel);
@@ -34,9 +20,5 @@ int main() {
     for (const auto &item: models) {
         std::cout << "Model Name: " << item.name() << std::endl;
     }
-
-    std::sort(models.begin(), models.end(), CompareModelInfoByLoadedFirst);
-    auto model = models.begin();
-    std::cout << "Chosen Model: " << model->name() << std::endl;
     return 0;
 }

Clients/Java/sample/src/main/kotlin/gay/pizza/stable/diffusion/sample/main.kt

@@ -49,8 +49,7 @@ fun main(args: Array<String>) {
   }
   println("available models:")
   for (model in modelListResponse.availableModelsList) {
-    val maybeLoadedComputeUnits = if (model.isLoaded) " loaded_compute_units=${model.loadedComputeUnits.name}" else ""
-    println("  model ${model.name} attention=${model.attention} loaded=${model.isLoaded}${maybeLoadedComputeUnits}")
+    println("  model ${model.name} attention=${model.attention}")
   }
 
   val model = if (chosenModelName == null) {
@@ -59,15 +58,11 @@ fun main(args: Array<String>) {
     modelListResponse.availableModelsList.first { it.name == chosenModelName }
   }
 
-  if (!model.isLoaded) {
-    println("loading model ${model.name}...")
-    client.modelServiceBlocking.loadModel(LoadModelRequest.newBuilder().apply {
-      modelName = model.name
-      computeUnits = model.supportedComputeUnitsList.first()
-    }.build())
-  } else {
-    println("using model ${model.name}...")
-  }
+  println("loading model ${model.name}...")
+  client.hostModelServiceBlocking.loadModel(LoadModelRequest.newBuilder().apply {
+    modelName = model.name
+    computeUnits = model.supportedComputeUnitsList.first()
+  }.build())
 
   println("tokenizing prompts...")
 

Clients/Java/src/main/kotlin/gay/pizza/stable/diffusion/StableDiffusionRpcClient.kt

@@ -20,6 +20,22 @@ class StableDiffusionRpcClient(val channel: Channel) {
     ModelServiceGrpcKt.ModelServiceCoroutineStub(channel)
   }
 
+  val hostModelService: HostModelServiceGrpc.HostModelServiceStub by lazy {
+    HostModelServiceGrpc.newStub(channel)
+  }
+
+  val hostModelServiceBlocking: HostModelServiceGrpc.HostModelServiceBlockingStub by lazy {
+    HostModelServiceGrpc.newBlockingStub(channel)
+  }
+
+  val hostModelServiceFuture: HostModelServiceGrpc.HostModelServiceFutureStub by lazy {
+    HostModelServiceGrpc.newFutureStub(channel)
+  }
+
+  val hostModelServiceCoroutine: HostModelServiceGrpcKt.HostModelServiceCoroutineStub by lazy {
+    HostModelServiceGrpcKt.HostModelServiceCoroutineStub(channel)
+  }
+
   val imageGenerationService: ImageGenerationServiceGrpc.ImageGenerationServiceStub by lazy {
     ImageGenerationServiceGrpc.newStub(channel)
   }