[feat] Add SGLANG_TOOL_STRICT_LEVEL for tool-call behavior control (#12423)

Signed-off-by: Xinyuan Tong <xinyuantong.cs@gmail.com>

python/sglang/srt/environ.py

@@ -2,6 +2,7 @@ import os
 import subprocess
 import warnings
 from contextlib import ExitStack, contextmanager
+from enum import IntEnum
 from typing import Any
 
 
@@ -105,6 +106,20 @@ class EnvFloat(EnvField):
             raise ValueError(f'"{value}" is not a valid float value')
 
 
+class ToolStrictLevel(IntEnum):
+    """
+    Defines the strictness levels for tool call parsing and validation.
+
+    OFF: No strict validation
+    FUNCTION: Enables structural tag constraints for all tools
+    PARAMETER: Enforces strict parameter validation for all tools
+    """
+
+    OFF = 0
+    FUNCTION = 1
+    PARAMETER = 2
+
+
 class Envs:
     # fmt: off
 
@@ -259,6 +274,9 @@ class Envs:
     # Sparse Embeddings
     SGLANG_EMBEDDINGS_SPARSE_HEAD = EnvStr(None)
 
+    # Tool-Call behavior
+    SGLANG_TOOL_STRICT_LEVEL = EnvInt(ToolStrictLevel.OFF)
+
     # fmt: on
 
 

python/sglang/srt/function_call/function_call_parser.py

@@ -8,6 +8,7 @@ from sglang.srt.entrypoints.openai.protocol import (
     ToolCallConstraint,
     ToolChoice,
 )
+from sglang.srt.environ import ToolStrictLevel, envs
 from sglang.srt.function_call.base_format_detector import BaseFormatDetector
 from sglang.srt.function_call.core_types import ToolCallItem
 from sglang.srt.function_call.deepseekv3_detector import DeepSeekV3Detector
@@ -62,6 +63,7 @@ class FunctionCallParser:
 
         self.detector = detector
         self.tools = tools
+        self.tool_strict_level = envs.SGLANG_TOOL_STRICT_LEVEL.get()
 
     def has_tool_call(self, text: str) -> bool:
         """
@@ -142,7 +144,10 @@ class FunctionCallParser:
             info = get_structure_info(name)
 
             # accept all if not strict, otherwise only accept the schema
-            schema = function.parameters if function.strict else {}
+            is_strict = (
+                function.strict or self.tool_strict_level >= ToolStrictLevel.PARAMETER
+            )
+            schema = function.parameters if is_strict else {}
 
             tool_structures.append(
                 StructuresResponseFormat(
@@ -180,7 +185,10 @@ class FunctionCallParser:
         if (
             self.detector.supports_structural_tag()
             and tool_choice == "auto"
-            and any(tool.function.strict for tool in self.tools)
+            and (
+                any(tool.function.strict for tool in self.tools)
+                or self.tool_strict_level >= ToolStrictLevel.FUNCTION
+            )
         ):
             tag = self.get_structure_tag()
             return ("structural_tag", tag)

python/sglang/srt/server_args.py

@@ -27,7 +27,7 @@ from typing import Dict, List, Literal, Optional, Union
 import orjson
 
 from sglang.srt.connector import ConnectorType
-from sglang.srt.environ import envs
+from sglang.srt.environ import ToolStrictLevel, envs
 from sglang.srt.function_call.function_call_parser import FunctionCallParser
 from sglang.srt.lora.lora_registry import LoRARef
 from sglang.srt.parser.reasoning_parser import ReasoningParser
@@ -1686,6 +1686,9 @@ class ServerArgs:
         os.environ["SGLANG_ENABLE_DETERMINISTIC_INFERENCE"] = (
             "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":
+            envs.SGLANG_TOOL_STRICT_LEVEL.set(ToolStrictLevel.PARAMETER)
 
     def _handle_cache_compatibility(self):
         if self.enable_hierarchical_cache and self.disable_radix_cache: