Set self.prev_tool_call_arr after generating the first DeltaToolCall

Signed-off-by: avigny <[email protected]>

Author: avigny

tests/tool_use/test_mistral_tool_parser.py

@@ -302,6 +302,8 @@ def _test_extract_tool_calls_streaming(tool_parser, tokenizer, model_output,
             assert len(streamed_tool_calls) == 1
             tool_call = streamed_tool_calls[0]
 
+            assert len(tool_parser.prev_tool_call_arr) > 0
+
             # if a new tool is being called, set up empty arguments
             if tool_call.index != tool_call_idx:
                 tool_call_idx = tool_call.index

vllm/entrypoints/openai/tool_parsers/mistral_tool_parser.py

@@ -245,6 +245,13 @@ def _extract_tool_calls_streaming(
             content=additional_content,
             tool_calls=delta_tool_calls,
         )
+        # HACK: serving_chat.py inspects the internal state of tool parsers
+        # when determining it's final streaming delta, automatically
+        # adding autocompleted JSON.
+        # These two lines avoid that nonsense while ensuring finish_reason
+        # is set to tool_calls when at least one tool is called.
+        if delta and not self.prev_tool_call_arr:
+            self.prev_tool_call_arr = [{"arguments": {}}]
         return delta
 
     def _generate_delta_tool_call(self,
@@ -476,6 +483,15 @@ def _extract_tool_calls_streaming_pre_v11_tokenizer(
 
         if current_tool_call_modified:
             delta_tool_calls.append(current_tool_call)
+
+        # HACK: serving_chat.py inspects the internal state of tool parsers
+        # when determining it's final streaming delta, automatically
+        # adding autocompleted JSON.
+        # These two lines avoid that nonsense while ensuring finish_reason
+        # is set to tool_calls when at least one tool is called.
+        if delta_tool_calls and not self.prev_tool_call_arr:
+            self.prev_tool_call_arr = [{"arguments": {}}]
+
         return DeltaMessage(content=content, tool_calls=delta_tool_calls)
 
     def _split_delta(