From 9409c43593f2d6d64595981abf216a15752b0875 Mon Sep 17 00:00:00 2001 From: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com> Date: Wed, 28 Jan 2026 23:58:16 -0500 Subject: [PATCH] Fix flaky tool calls in the Kimi K2.5 model (#17914) Signed-off-by: Xinyuan Tong --- python/sglang/srt/server_args.py | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py index 66750a5d3..49975de64 100644 --- a/python/sglang/srt/server_args.py +++ b/python/sglang/srt/server_args.py @@ -27,7 +27,7 @@ import tempfile from typing import Any, Callable, Dict, List, Literal, Optional, Union from sglang.srt.connector import ConnectorType -from sglang.srt.environ import ToolStrictLevel, envs +from sglang.srt.environ import envs from sglang.srt.function_call.function_call_parser import FunctionCallParser from sglang.srt.layers.attention.fla.chunk_delta_h import CHUNK_SIZE as FLA_CHUNK_SIZE from sglang.srt.lora.lora_registry import LoRARef @@ -2485,12 +2485,6 @@ class ServerArgs: envs.SGLANG_ENABLE_DETERMINISTIC_INFERENCE.set( "1" if self.enable_deterministic_inference else "0" ) - # Set the highest strict level for Kimi K2 tool calls - if ( - self.tool_call_parser == "kimi_k2" - and not envs.SGLANG_TOOL_STRICT_LEVEL.is_set() - ): - envs.SGLANG_TOOL_STRICT_LEVEL.set(ToolStrictLevel.PARAMETER) def _handle_cache_compatibility(self): if self.enable_hierarchical_cache and self.disable_radix_cache: