openapi: finish streaming tool calls as tool_calls

Author: anuraaga

openai/openai.go

@@ -266,6 +266,9 @@ func toChunk(id string, r api.ChatResponse) ChatCompletionChunk {
 			Index: 0,
 			Delta: Message{Role: "assistant", Content: r.Message.Content, ToolCalls: toolCalls},
 			FinishReason: func(reason string) *string {
+				if len(toolCalls) > 0 {
+					reason = "tool_calls"
+				}
 				if len(reason) > 0 {
 					return &reason
 				}
@@ -570,8 +573,9 @@ type BaseWriter struct {
 }
 
 type ChatWriter struct {
-	stream bool
-	id     string
+	stream   bool
+	finished bool
+	id       string
 	BaseWriter
 }
 
@@ -620,7 +624,15 @@ func (w *ChatWriter) writeResponse(data []byte) (int, error) {
 
 	// chat chunk
 	if w.stream {
-		d, err := json.Marshal(toChunk(w.id, chatResponse))
+		chunk := toChunk(w.id, chatResponse)
+		if w.finished {
+			// we've already finished the chat, usually for tool calls, so do
+			// not continue to send empty choices.
+			chunk.Choices = nil
+		} else if len(chunk.Choices) > 0 && chunk.Choices[0].FinishReason != nil {
+			w.finished = true
+		}
+		d, err := json.Marshal(chunk)
 		if err != nil {
 			return 0, err
 		}