diff --git a/core/backend/llm.go b/core/backend/llm.go
index 9e121f79..9a4d0d46 100644
--- a/core/backend/llm.go
+++ b/core/backend/llm.go
@@ -118,7 +118,7 @@ func ModelInference(ctx context.Context, s string, messages []schema.Message, im
 
 			var partialRune []byte
 			err := inferenceModel.PredictStream(ctx, opts, func(reply *proto.Reply) {
-				msg := reply.GetMessage()
+				msg := reply.Message
 				partialRune = append(partialRune, msg...)
 
 				tokenUsage.Prompt = int(reply.PromptTokens)
diff --git a/core/http/endpoints/openai/chat.go b/core/http/endpoints/openai/chat.go
index b03b18bd..1ac1387e 100644
--- a/core/http/endpoints/openai/chat.go
+++ b/core/http/endpoints/openai/chat.go
@@ -39,15 +39,11 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, startup
 		responses <- initialMessage
 
 		ComputeChoices(req, s, config, startupOptions, loader, func(s string, c *[]schema.Choice) {}, func(s string, usage backend.TokenUsage) bool {
-			choices := []schema.Choice{}
-			if s != "" {
-				choices = append(choices, schema.Choice{Delta: &schema.Message{Content: &s}, Index: 0})
-			}
 			resp := schema.OpenAIResponse{
 				ID:      id,
 				Created: created,
 				Model:   req.Model, // we have to return what the user sent here, due to OpenAI spec.
-				Choices: choices,
+				Choices: []schema.Choice{{Delta: &schema.Message{Content: &s}, Index: 0}},
 				Object:  "chat.completion.chunk",
 				Usage: schema.OpenAIUsage{
 					PromptTokens:     usage.Prompt,
@@ -469,9 +465,6 @@ func ChatEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, startup
 				toolsCalled := false
 				for ev := range responses {
 					usage = &ev.Usage // Copy a pointer to the latest usage chunk so that the stop message can reference it
-					if len(ev.Choices) == 0 {
-						break
-					}
 					if len(ev.Choices[0].Delta.ToolCalls) > 0 {
 						toolsCalled = true
 					}