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, 01:15 UTC
[Bugfix] Interleaved thinking keeps compatibility with reasoning_content (#33635)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com> Co-authored-by: Koushik Dutta <koushd@gmail.com>
Chauncey committed
Feb 3, 2026 at 14:46 UTC
bf001da4bfb53854927b68055a12efd05d494786
1 file changed
+3
vllm/entrypoints/chat_utils.py
+3
@@ -1463,6 +1463,9 @@ def _parse_chat_message_content(
1463
# Include reasoning if present for interleaved thinking.
1464
if reasoning is not None:
1465
result_msg["reasoning"] = cast(str, reasoning)
1466
+ result_msg["reasoning_content"] = cast(
1467
+ str, reasoning
1468
+ ) # keep compatibility
1469
elif role == "tool":
1470
parsed_msg = _ToolParser(message)
1471
if "tool_call_id" in parsed_msg: