From bf9a1853958584fe039d33242a43c91cf8786d61 Mon Sep 17 00:00:00 2001
From: Robin Nabel <opensource@nabel.co>
Date: Mon, 16 Mar 2026 11:48:52 +0100
Subject: [PATCH] GLM4 tool parser: fix streaming mode (#35208)

Signed-off-by: Robin Nabel <opensource@nabel.co>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>
---
 .../tool_parsers/test_glm4_moe_tool_parser.py | 24 ++++++++++++++-----
 vllm/tool_parsers/glm4_moe_tool_parser.py     | 12 ++++++----
 2 files changed, 26 insertions(+), 10 deletions(-)

diff --git a/tests/tool_parsers/test_glm4_moe_tool_parser.py b/tests/tool_parsers/test_glm4_moe_tool_parser.py
index 292714cde..9ee9ea008 100644
--- a/tests/tool_parsers/test_glm4_moe_tool_parser.py
+++ b/tests/tool_parsers/test_glm4_moe_tool_parser.py
@@ -560,19 +560,23 @@ def test_streaming_empty_tool_call(glm4_moe_tool_parser, mock_request):
     assert glm4_moe_tool_parser.current_tool_id == -1
 
 
-def test_streaming_prev_tool_call_arr_finalization(glm4_moe_tool_parser, mock_request):
+def test_streaming_prev_tool_call_arr_updates(glm4_moe_tool_parser, mock_request):
     """Test that prev_tool_call_arr contains parsed dict after tool call."""
     _reset_streaming_state(glm4_moe_tool_parser)
 
     # Stream a complete tool call
+    name_only = {"name": "get_weather", "arguments": {}}
+    name_and_args = {"name": "get_weather", "arguments": {"city": "Beijing"}}
     chunks = [
-        "<tool_call>get_weather\n",
-        "<arg_key>city</arg_key>",
-        "<arg_value>Beijing</arg_value>",
-        "</tool_call>",
+        # Delta, expected streamed_args_for_tool, expected prev_tool_call_arr
+        ("<tool_call>get_weather\n", "", name_only),
+        ("<arg_key>city</arg_key>", "", name_only),
+        ("<arg_value>Beijing</arg_value>", '{"city": "Beijing"', name_only),
+        # Note: arguments are only updated when the tool call is complete.
+        ("</tool_call>", '{"city": "Beijing"}', name_and_args),
     ]
 
-    for chunk in chunks:
+    for chunk, exp_streamed, exp_prev_tc in chunks:
         glm4_moe_tool_parser.extract_tool_calls_streaming(
             previous_text="",
             current_text="",
@@ -582,6 +586,8 @@ def test_streaming_prev_tool_call_arr_finalization(glm4_moe_tool_parser, mock_re
             delta_token_ids=[],
             request=mock_request,
         )
+        assert glm4_moe_tool_parser.streamed_args_for_tool[0] == exp_streamed
+        assert glm4_moe_tool_parser.prev_tool_call_arr[0] == exp_prev_tc
 
     # After the tool call completes, prev_tool_call_arr should have parsed dict
     assert len(glm4_moe_tool_parser.prev_tool_call_arr) == 1
@@ -592,6 +598,12 @@ def test_streaming_prev_tool_call_arr_finalization(glm4_moe_tool_parser, mock_re
     assert isinstance(args, dict), f"Expected dict, got {type(args)}"
     assert args.get("city") == "Beijing"
 
+    # Test equivalence of prev_tool_call_arr and streamed_args_for_tool
+    # Simulates logic in chat_completion/serving.py:chat_completion_stream_generator
+    tool_call_json = json.dumps(tool_entry.get("arguments", {}))
+    streamed_content = glm4_moe_tool_parser.streamed_args_for_tool[0]
+    assert tool_call_json.startswith(streamed_content)
+
 
 def test_streaming_multiple_tool_calls_sequential(glm4_moe_tool_parser, mock_request):
     """Test streaming multiple sequential tool calls."""
diff --git a/vllm/tool_parsers/glm4_moe_tool_parser.py b/vllm/tool_parsers/glm4_moe_tool_parser.py
index d6942e854..2a03c8583 100644
--- a/vllm/tool_parsers/glm4_moe_tool_parser.py
+++ b/vllm/tool_parsers/glm4_moe_tool_parser.py
@@ -337,10 +337,10 @@ class Glm4MoeModelToolParser(ToolParser):
                     key_json = json.dumps(key, ensure_ascii=False)
 
                     if not self._args_started[self.current_tool_id]:
-                        frag = "{" + key_json + ':"'
+                        frag = "{" + key_json + ': "'
                         self._args_started[self.current_tool_id] = True
                     else:
-                        frag = "," + key_json + ':"'
+                        frag = ", " + key_json + ': "'
 
                     self.streamed_args_for_tool[self.current_tool_id] += frag
                     self._streaming_string_value = True
@@ -447,6 +447,10 @@ class Glm4MoeModelToolParser(ToolParser):
         self.current_tool_id -= 1
 
     def _emit_tool_name_delta(self, tool_name: str) -> DeltaMessage:
+        self.prev_tool_call_arr[self.current_tool_id] = {
+            "name": self._current_tool_name,
+            "arguments": {},
+        }
         return DeltaMessage(
             tool_calls=[
                 DeltaToolCall(
@@ -493,10 +497,10 @@ class Glm4MoeModelToolParser(ToolParser):
         val_json = json.dumps(val_obj, ensure_ascii=False)
 
         if not self._args_started[self.current_tool_id]:
-            fragment = "{" + key_json + ":" + val_json
+            fragment = "{" + key_json + ": " + val_json
             self._args_started[self.current_tool_id] = True
         else:
-            fragment = "," + key_json + ":" + val_json
+            fragment = "," + key_json + ": " + val_json
 
         self._seen_keys[self.current_tool_id].add(key)
         self.streamed_args_for_tool[self.current_tool_id] += fragment