[Fix] [gpt-oss] fix non-tool calling path for chat completion

vllm/entrypoints/openai/serving_chat.py

@@ -90,6 +90,11 @@ def __init__(
         self.chat_template = chat_template
         self.chat_template_content_format: Final = chat_template_content_format
         self.enable_log_outputs = enable_log_outputs
+        self.use_harmony = model_config.hf_config.model_type == "gpt_oss"
+        if self.use_harmony:
+            # override the default tool_parser and auto tools for harmony
+            tool_parser = 'openai'
+            enable_auto_tools = True
 
         # set up tool use
         self.enable_auto_tools: bool = enable_auto_tools
@@ -141,7 +146,6 @@ def __init__(
         else:
             self.tool_call_id_type = 'random'
 
-        self.use_harmony = model_config.hf_config.model_type == "gpt_oss"
         if self.use_harmony:
             if "stop_token_ids" not in self.default_sampling_params:
                 self.default_sampling_params["stop_token_ids"] = []