A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 10:00 UTC
Fix tool call indexing double-counting (#33141)
Signed-off-by: wanglinian <wanglinian@stu.pku.edu.cn>
wangln19 committed
Jan 29, 2026 at 13:57 UTC
39037d258e68da3926d99681ea63e46212e519f9
1 file changed
+4
-3
vllm/entrypoints/openai/chat_completion/serving.py
+4
-3
@@ -982,6 +982,7 @@ class OpenAIServingChat(OpenAIServing):
982
index=i,
983
)
984
function_name_returned[i] = True
985
+ history_tool_call_cnt += 1
986
987
delta_message = DeltaMessage(
988
tool_calls=[
@@ -1582,7 +1583,7 @@ class OpenAIServingChat(OpenAIServing):
1583
generated_id = make_tool_call_id(
1584
id_type=self.tool_call_id_type,
1585
func_name=tc.name,
1585
- idx=history_tool_call_cnt + idx,
1586
+ idx=history_tool_call_cnt,
1587
)
1588
tool_call_class_items.append(
1589
tool_call_class(id=generated_id, function=tc)
@@ -1617,7 +1618,7 @@ class OpenAIServingChat(OpenAIServing):
1618
generated_id = make_tool_call_id(
1619
id_type=self.tool_call_id_type,
1620
func_name=tool_call.name,
1620
- idx=history_tool_call_cnt + idx,
1621
+ idx=history_tool_call_cnt,
1622
)
1623
tool_call_class_items.append(
1624
tool_call_class(id=generated_id, function=tool_call)
@@ -1665,7 +1666,7 @@ class OpenAIServingChat(OpenAIServing):
1666
generated_id = make_tool_call_id(
1667
id_type=self.tool_call_id_type,
1668
func_name=tc.name,
1668
- idx=history_tool_call_cnt + idx,
1669
+ idx=history_tool_call_cnt,
1670
)
1671
tool_call_items.append(
1672
tool_call_class(id=generated_id, function=tc)