feat(grpc): backend SPI pluggable in embedding mode (#1621)

* run server

* grpc backend embedded support

* backend providable

api/backend/embeddings.go

@@ -41,7 +41,7 @@ func ModelEmbedding(s string, tokens []int, loader *model.ModelLoader, c config.
 
 	var fn func() ([]float32, error)
 	switch model := inferenceModel.(type) {
-	case *grpc.Client:
+	case grpc.Backend:
 		fn = func() ([]float32, error) {
 			predictOptions := gRPCPredictOpts(c, loader.ModelPath)
 			if len(tokens) > 0 {

api/backend/llm.go

@@ -31,7 +31,7 @@ func ModelInference(ctx context.Context, s string, images []string, loader *mode
 
 	grpcOpts := gRPCModelOpts(c)
 
-	var inferenceModel *grpc.Client
+	var inferenceModel grpc.Backend
 	var err error
 
 	opts := modelOpts(c, o, []model.Option{