[Refactor] [2/N] Move tool parsers into the vLLM main directory (#30675)

Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>

docs/features/tool_calling.md

@@ -420,7 +420,7 @@ Flags: `--tool-call-parser pythonic --chat-template {see_above}`
 
 ## How to Write a Tool Parser Plugin
 
-A tool parser plugin is a Python file containing one or more ToolParser implementations. You can write a ToolParser similar to the `Hermes2ProToolParser` in [vllm/entrypoints/openai/tool_parsers/hermes_tool_parser.py](../../vllm/entrypoints/openai/tool_parsers/hermes_tool_parser.py).
+A tool parser plugin is a Python file containing one or more ToolParser implementations. You can write a ToolParser similar to the `Hermes2ProToolParser` in [vllm/tool_parsers/hermes_tool_parser.py](../../vllm/tool_parsers/hermes_tool_parser.py).
 
 Here is a summary of a plugin file:
 
@@ -468,7 +468,7 @@ Here is a summary of a plugin file:
     # register the tool parser to ToolParserManager
     ToolParserManager.register_lazy_module(
         name="example",
-        module_path="vllm.entrypoints.openai.tool_parsers.example",
+        module_path="vllm.tool_parsers.example",
         class_name="ExampleToolParser",
     )
 

tests/entrypoints/openai/test_serving_chat.py

@@ -19,9 +19,9 @@ from vllm.entrypoints.openai.protocol import (
 )
 from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 from vllm.entrypoints.openai.serving_models import BaseModelPath, OpenAIServingModels
-from vllm.entrypoints.openai.tool_parsers import ToolParserManager
 from vllm.outputs import CompletionOutput, RequestOutput
 from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers import ToolParserManager
 from vllm.v1.engine.async_llm import AsyncLLM
 
 from ...utils import RemoteOpenAIServer

tests/entrypoints/openai/tool_parsers/test_gigachat3_tool_parser.py

@@ -10,8 +10,8 @@ from tests.entrypoints.openai.tool_parsers.utils import (
     run_tool_extraction_streaming,
 )
 from vllm.entrypoints.openai.protocol import FunctionCall
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers import ToolParser, ToolParserManager
 
 SIMPLE_ARGS_DICT = {
     "action": "create",

tests/entrypoints/openai/tool_parsers/test_hermes_tool_parser.py

@@ -6,8 +6,8 @@ import json
 import pytest
 
 from vllm.entrypoints.openai.protocol import ChatCompletionRequest
-from vllm.entrypoints.openai.tool_parsers.hermes_tool_parser import Hermes2ProToolParser
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.hermes_tool_parser import Hermes2ProToolParser
 
 from ....utils import RemoteOpenAIServer
 

tests/entrypoints/openai/tool_parsers/test_hunyuan_a13b_tool_parser.py

@@ -12,7 +12,7 @@ from tests.entrypoints.openai.tool_parsers.utils import (
     run_tool_extraction_streaming,
 )
 from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
+from vllm.tool_parsers import ToolParser, ToolParserManager
 
 
 def make_tool_call(name, arguments):

tests/entrypoints/openai/tool_parsers/test_llama3_json_tool_parser.py

@@ -6,8 +6,8 @@ from unittest.mock import MagicMock, patch
 import pytest
 
 from vllm.entrypoints.openai.protocol import ExtractedToolCallInformation
-from vllm.entrypoints.openai.tool_parsers.llama_tool_parser import Llama3JsonToolParser
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.llama_tool_parser import Llama3JsonToolParser
 
 
 @pytest.fixture

tests/entrypoints/openai/tool_parsers/test_llama4_pythonic_tool_parser.py

@@ -10,8 +10,8 @@ from tests.entrypoints.openai.tool_parsers.utils import (
     run_tool_extraction_streaming,
 )
 from vllm.entrypoints.openai.protocol import FunctionCall
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers import ToolParser, ToolParserManager
 
 # Test cases similar to pythonic parser but with Llama4 specific format
 SIMPLE_FUNCTION_OUTPUT = "[get_weather(city='LA', metric='C')]"

tests/entrypoints/openai/tool_parsers/test_olmo3_tool_parser.py

@@ -10,8 +10,8 @@ from tests.entrypoints.openai.tool_parsers.utils import (
     run_tool_extraction_streaming,
 )
 from vllm.entrypoints.openai.protocol import FunctionCall
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers import ToolParser, ToolParserManager
 
 # https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/text_prompt_format.md#model-response-format-1
 SIMPLE_FUNCTION_OUTPUT = "get_weather(city='San Francisco', metric='celsius')"

tests/entrypoints/openai/tool_parsers/test_pythonic_tool_parser.py

@@ -10,8 +10,8 @@ from tests.entrypoints.openai.tool_parsers.utils import (
     run_tool_extraction_streaming,
 )
 from vllm.entrypoints.openai.protocol import FunctionCall
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers import ToolParser, ToolParserManager
 
 # https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/text_prompt_format.md#model-response-format-1
 SIMPLE_FUNCTION_OUTPUT = "get_weather(city='San Francisco', metric='celsius')"

tests/entrypoints/openai/tool_parsers/utils.py

@@ -10,8 +10,8 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers import ToolParser
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers import ToolParser
 
 
 class StreamingToolReconstructor:

tests/models/language/generation/test_mistral.py

@@ -5,12 +5,12 @@ import json
 
 import pytest
 
-from vllm.entrypoints.openai.tool_parsers.mistral_tool_parser import (
+from vllm.sampling_params import SamplingParams
+from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers.mistral_tool_parser import (
     MistralToolCall,
     MistralToolParser,
 )
-from vllm.sampling_params import SamplingParams
-from vllm.tokenizers.mistral import MistralTokenizer
 
 from ...utils import check_logprobs_close
 

tests/tool_use/test_deepseekv31_tool_parser.py

@@ -3,10 +3,10 @@
 
 import pytest
 
-from vllm.entrypoints.openai.tool_parsers.deepseekv31_tool_parser import (
+from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers.deepseekv31_tool_parser import (
     DeepSeekV31ToolParser,
 )
-from vllm.tokenizers import get_tokenizer
 
 MODEL = "deepseek-ai/DeepSeek-V3.1"
 

tests/tool_use/test_ernie45_moe_tool_parser.py

@@ -13,9 +13,9 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.ernie45_tool_parser import Ernie45ToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tool_parsers.ernie45_tool_parser import Ernie45ToolParser
 
 # Use a common model that is likely to be available
 MODEL = "baidu/ERNIE-4.5-21B-A3B-Thinking"

tests/tool_use/test_glm4_moe_tool_parser.py

@@ -7,10 +7,10 @@ import json
 import pytest
 
 from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
-from vllm.entrypoints.openai.tool_parsers.glm4_moe_tool_parser import (
+from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers.glm4_moe_tool_parser import (
     Glm4MoeModelToolParser,
 )
-from vllm.tokenizers import get_tokenizer
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_jamba_tool_parser.py

@@ -9,9 +9,9 @@ import pytest
 from partial_json_parser.core.options import Allow
 
 from vllm.entrypoints.openai.protocol import DeltaMessage, FunctionCall, ToolCall
-from vllm.entrypoints.openai.tool_parsers.jamba_tool_parser import JambaToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tool_parsers.jamba_tool_parser import JambaToolParser
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_kimi_k2_tool_parser.py

@@ -7,8 +7,8 @@ import json
 import pytest
 
 from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
-from vllm.entrypoints.openai.tool_parsers.kimi_k2_tool_parser import KimiK2ToolParser
 from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers.kimi_k2_tool_parser import KimiK2ToolParser
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_minimax_tool_parser.py

@@ -12,8 +12,8 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.minimax_tool_parser import MinimaxToolParser
 from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers.minimax_tool_parser import MinimaxToolParser
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_mistral_tool_parser.py

@@ -12,10 +12,10 @@ from mistral_common.protocol.instruct.tool_calls import FunctionCall, ToolCall
 from partial_json_parser.core.options import Allow
 
 from vllm.entrypoints.openai.protocol import DeltaMessage, DeltaToolCall
-from vllm.entrypoints.openai.tool_parsers.mistral_tool_parser import MistralToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
 from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers.mistral_tool_parser import MistralToolParser
 
 
 @pytest.fixture(scope="module")

tests/tool_use/test_openai_tool_parser.py

@@ -15,8 +15,8 @@ from openai_harmony import (
 )
 
 from vllm.entrypoints.openai.protocol import FunctionCall, ToolCall
-from vllm.entrypoints.openai.tool_parsers.openai_tool_parser import OpenAIToolParser
 from vllm.tokenizers import get_tokenizer
+from vllm.tool_parsers.openai_tool_parser import OpenAIToolParser
 
 MODEL = "gpt2"
 

tests/tool_use/test_qwen3coder_tool_parser.py

@@ -13,12 +13,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.qwen3coder_tool_parser import (
-    Qwen3CoderToolParser,
-)
-from vllm.entrypoints.openai.tool_parsers.qwen3xml_tool_parser import Qwen3XMLToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tool_parsers.qwen3coder_tool_parser import (
+    Qwen3CoderToolParser,
+)
+from vllm.tool_parsers.qwen3xml_tool_parser import Qwen3XMLToolParser
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_seed_oss_tool_parser.py

@@ -14,9 +14,9 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.seed_oss_tool_parser import SeedOssToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tool_parsers.seed_oss_tool_parser import SeedOssToolParser
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_tool_choice_required.py

@@ -12,7 +12,7 @@ from vllm.entrypoints.openai.protocol import (
     ChatCompletionToolsParam,
 )
 from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
-from vllm.entrypoints.openai.tool_parsers.utils import get_json_schema_from_tools
+from vllm.tool_parsers.utils import get_json_schema_from_tools
 
 pytestmark = pytest.mark.cpu_test
 

tests/tool_use/test_xlam_tool_parser.py

@@ -12,9 +12,9 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.xlam_tool_parser import xLAMToolParser
 from vllm.tokenizers import TokenizerLike, get_tokenizer
 from vllm.tokenizers.detokenizer_utils import detokenize_incrementally
+from vllm.tool_parsers.xlam_tool_parser import xLAMToolParser
 
 pytestmark = pytest.mark.cpu_test
 

vllm/entrypoints/context.py

@@ -34,13 +34,13 @@ from vllm.entrypoints.openai.protocol import (
     ResponseRawMessageAndToken,
     ResponsesRequest,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import ToolParser
 from vllm.entrypoints.responses_utils import construct_tool_dicts
 from vllm.entrypoints.tool import Tool
 from vllm.entrypoints.tool_server import ToolServer
 from vllm.outputs import RequestOutput
 from vllm.reasoning.abs_reasoning_parsers import ReasoningParser
 from vllm.tokenizers.protocol import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import ToolParser
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 from vllm.utils import random_uuid
 

vllm/entrypoints/openai/api_server.py

@@ -72,7 +72,6 @@ from vllm.entrypoints.openai.serving_transcription import (
     OpenAIServingTranscription,
     OpenAIServingTranslation,
 )
-from vllm.entrypoints.openai.tool_parsers import ToolParserManager
 from vllm.entrypoints.openai.utils import validate_json_request
 from vllm.entrypoints.pooling.classify.serving import ServingClassification
 from vllm.entrypoints.pooling.embed.serving import OpenAIServingEmbedding
@@ -95,6 +94,7 @@ from vllm.entrypoints.utils import (
 from vllm.logger import init_logger
 from vllm.reasoning import ReasoningParserManager
 from vllm.tasks import POOLING_TASKS
+from vllm.tool_parsers import ToolParserManager
 from vllm.usage.usage_lib import UsageContext
 from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.gc_utils import freeze_gc_heap

vllm/entrypoints/openai/cli_args.py

@@ -27,8 +27,8 @@ from vllm.entrypoints.constants import (
     H11_MAX_INCOMPLETE_EVENT_SIZE_DEFAULT,
 )
 from vllm.entrypoints.openai.serving_models import LoRAModulePath
-from vllm.entrypoints.openai.tool_parsers import ToolParserManager
 from vllm.logger import init_logger
+from vllm.tool_parsers import ToolParserManager
 from vllm.utils.argparse_utils import FlexibleArgumentParser
 
 logger = init_logger(__name__)

vllm/entrypoints/openai/parser/responses_parser.py

@@ -12,10 +12,10 @@ from openai.types.responses.response_reasoning_item import (
 )
 
 from vllm.entrypoints.openai.protocol import ResponseInputOutputItem, ResponsesRequest
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import ToolParser
 from vllm.outputs import CompletionOutput
 from vllm.reasoning.abs_reasoning_parsers import ReasoningParser
 from vllm.tokenizers.protocol import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import ToolParser
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 from vllm.utils import random_uuid
 

vllm/entrypoints/openai/serving_chat.py

@@ -57,8 +57,6 @@ from vllm.entrypoints.openai.serving_engine import (
     clamp_prompt_logprobs,
 )
 from vllm.entrypoints.openai.serving_models import OpenAIServingModels
-from vllm.entrypoints.openai.tool_parsers import ToolParser
-from vllm.entrypoints.openai.tool_parsers.mistral_tool_parser import MistralToolCall
 from vllm.entrypoints.openai.utils import maybe_filter_parallel_tool_calls
 from vllm.entrypoints.utils import get_max_tokens, should_include_usage
 from vllm.inputs.data import TokensPrompt
@@ -73,6 +71,8 @@ from vllm.tokenizers.mistral import (
     truncate_tool_call_ids,
     validate_request_params,
 )
+from vllm.tool_parsers import ToolParser
+from vllm.tool_parsers.mistral_tool_parser import MistralToolCall
 from vllm.utils.collection_utils import as_list
 from vllm.v1.sample.logits_processor import validate_logits_processors_parameters
 

vllm/entrypoints/openai/serving_engine.py

@@ -59,7 +59,6 @@ from vllm.entrypoints.openai.protocol import (
     TranslationRequest,
 )
 from vllm.entrypoints.openai.serving_models import OpenAIServingModels
-from vllm.entrypoints.openai.tool_parsers import ToolParser, ToolParserManager
 from vllm.entrypoints.pooling.classify.protocol import (
     ClassificationChatRequest,
     ClassificationCompletionRequest,
@@ -104,6 +103,7 @@ from vllm.sampling_params import BeamSearchParams, SamplingParams
 from vllm.tokenizers import TokenizerLike
 from vllm.tokenizers.deepseek_v32 import DeepseekV32Tokenizer
 from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers import ToolParser, ToolParserManager
 from vllm.tracing import (
     contains_trace_headers,
     extract_trace_headers,

vllm/entrypoints/openai/tool_parsers/__init__.py

@@ -1,150 +1,33 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-    ToolParserManager,
-)
-
-__all__ = ["ToolParser", "ToolParserManager"]
+import warnings
 
 
-"""
-Register a lazy module mapping.
+def __getattr__(name: str):
+    if name == "ToolParser":
+        from vllm.tool_parsers import ToolParser
 
-Example:
-    ToolParserManager.register_lazy_module(
-        name="kimi_k2",
-        module_path="vllm.entrypoints.openai.tool_parsers.kimi_k2_parser",
-        class_name="KimiK2ToolParser",
-    )
-"""
+        warnings.warn(
+            "`vllm.entrypoints.openai.tool_parsers.ToolParser` has been moved to "
+            "`vllm.tool_parsers.ToolParser`. "
+            "The old name will be removed in v0.14.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
 
+        return ToolParser
+    if name == "ToolParserManager":
+        from vllm.tool_parsers import ToolParserManager
 
-_TOOL_PARSERS_TO_REGISTER = {
-    "deepseek_v3": (  # name
-        "deepseekv3_tool_parser",  # filename
-        "DeepSeekV3ToolParser",  # class_name
-    ),
-    "deepseek_v31": (
-        "deepseekv31_tool_parser",
-        "DeepSeekV31ToolParser",
-    ),
-    "deepseek_v32": (
-        "deepseekv32_tool_parser",
-        "DeepSeekV32ToolParser",
-    ),
-    "ernie45": (
-        "ernie45_tool_parser",
-        "Ernie45ToolParser",
-    ),
-    "glm45": (
-        "glm4_moe_tool_parser",
-        "Glm4MoeModelToolParser",
-    ),
-    "granite-20b-fc": (
-        "granite_20b_fc_tool_parser",
-        "Granite20bFCToolParser",
-    ),
-    "granite": (
-        "granite_tool_parser",
-        "GraniteToolParser",
-    ),
-    "hermes": (
-        "hermes_tool_parser",
-        "Hermes2ProToolParser",
-    ),
-    "hunyuan_a13b": (
-        "hunyuan_a13b_tool_parser",
-        "HunyuanA13BToolParser",
-    ),
-    "internlm": (
-        "internlm2_tool_parser",
-        "Internlm2ToolParser",
-    ),
-    "jamba": (
-        "jamba_tool_parser",
-        "JambaToolParser",
-    ),
-    "kimi_k2": (
-        "kimi_k2_tool_parser",
-        "KimiK2ToolParser",
-    ),
-    "llama3_json": (
-        "llama_tool_parser",
-        "Llama3JsonToolParser",
-    ),
-    "llama4_json": (
-        "llama_tool_parser",
-        "Llama3JsonToolParser",
-    ),
-    "llama4_pythonic": (
-        "llama4_pythonic_tool_parser",
-        "Llama4PythonicToolParser",
-    ),
-    "longcat": (
-        "longcat_tool_parser",
-        "LongcatFlashToolParser",
-    ),
-    "minimax_m2": (
-        "minimax_m2_tool_parser",
-        "MinimaxM2ToolParser",
-    ),
-    "minimax": (
-        "minimax_tool_parser",
-        "MinimaxToolParser",
-    ),
-    "mistral": (
-        "mistral_tool_parser",
-        "MistralToolParser",
-    ),
-    "olmo3": (
-        "olmo3_tool_parser",
-        "Olmo3PythonicToolParser",
-    ),
-    "openai": (
-        "openai_tool_parser",
-        "OpenAIToolParser",
-    ),
-    "phi4_mini_json": (
-        "phi4mini_tool_parser",
-        "Phi4MiniJsonToolParser",
-    ),
-    "pythonic": (
-        "pythonic_tool_parser",
-        "PythonicToolParser",
-    ),
-    "qwen3_coder": (
-        "qwen3coder_tool_parser",
-        "Qwen3CoderToolParser",
-    ),
-    "qwen3_xml": (
-        "qwen3xml_tool_parser",
-        "Qwen3XMLToolParser",
-    ),
-    "seed_oss": (
-        "seed_oss_tool_parser",
-        "SeedOssToolParser",
-    ),
-    "step3": (
-        "step3_tool_parser",
-        "Step3ToolParser",
-    ),
-    "xlam": (
-        "xlam_tool_parser",
-        "xLAMToolParser",
-    ),
-    "gigachat3": (
-        "gigachat3_tool_parser",
-        "GigaChat3ToolParser",
-    ),
-}
+        warnings.warn(
+            "`vllm.entrypoints.openai.tool_parsers.ToolParserManager` "
+            "has been moved to `vllm.tool_parsers.ToolParserManager`. "
+            "The old name will be removed in v0.14.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
 
+        return ToolParserManager
 
-def register_lazy_tool_parsers():
-    for name, (file_name, class_name) in _TOOL_PARSERS_TO_REGISTER.items():
-        module_path = f"vllm.entrypoints.openai.tool_parsers.{file_name}"
-        ToolParserManager.register_lazy_module(name, module_path, class_name)
-
-
-register_lazy_tool_parsers()
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

vllm/tool_parsers/__init__.py

@@ -0,0 +1,150 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+    ToolParserManager,
+)
+
+__all__ = ["ToolParser", "ToolParserManager"]
+
+
+"""
+Register a lazy module mapping.
+
+Example:
+    ToolParserManager.register_lazy_module(
+        name="kimi_k2",
+        module_path="vllm.tool_parsers.kimi_k2_parser",
+        class_name="KimiK2ToolParser",
+    )
+"""
+
+
+_TOOL_PARSERS_TO_REGISTER = {
+    "deepseek_v3": (  # name
+        "deepseekv3_tool_parser",  # filename
+        "DeepSeekV3ToolParser",  # class_name
+    ),
+    "deepseek_v31": (
+        "deepseekv31_tool_parser",
+        "DeepSeekV31ToolParser",
+    ),
+    "deepseek_v32": (
+        "deepseekv32_tool_parser",
+        "DeepSeekV32ToolParser",
+    ),
+    "ernie45": (
+        "ernie45_tool_parser",
+        "Ernie45ToolParser",
+    ),
+    "glm45": (
+        "glm4_moe_tool_parser",
+        "Glm4MoeModelToolParser",
+    ),
+    "granite-20b-fc": (
+        "granite_20b_fc_tool_parser",
+        "Granite20bFCToolParser",
+    ),
+    "granite": (
+        "granite_tool_parser",
+        "GraniteToolParser",
+    ),
+    "hermes": (
+        "hermes_tool_parser",
+        "Hermes2ProToolParser",
+    ),
+    "hunyuan_a13b": (
+        "hunyuan_a13b_tool_parser",
+        "HunyuanA13BToolParser",
+    ),
+    "internlm": (
+        "internlm2_tool_parser",
+        "Internlm2ToolParser",
+    ),
+    "jamba": (
+        "jamba_tool_parser",
+        "JambaToolParser",
+    ),
+    "kimi_k2": (
+        "kimi_k2_tool_parser",
+        "KimiK2ToolParser",
+    ),
+    "llama3_json": (
+        "llama_tool_parser",
+        "Llama3JsonToolParser",
+    ),
+    "llama4_json": (
+        "llama_tool_parser",
+        "Llama3JsonToolParser",
+    ),
+    "llama4_pythonic": (
+        "llama4_pythonic_tool_parser",
+        "Llama4PythonicToolParser",
+    ),
+    "longcat": (
+        "longcat_tool_parser",
+        "LongcatFlashToolParser",
+    ),
+    "minimax_m2": (
+        "minimax_m2_tool_parser",
+        "MinimaxM2ToolParser",
+    ),
+    "minimax": (
+        "minimax_tool_parser",
+        "MinimaxToolParser",
+    ),
+    "mistral": (
+        "mistral_tool_parser",
+        "MistralToolParser",
+    ),
+    "olmo3": (
+        "olmo3_tool_parser",
+        "Olmo3PythonicToolParser",
+    ),
+    "openai": (
+        "openai_tool_parser",
+        "OpenAIToolParser",
+    ),
+    "phi4_mini_json": (
+        "phi4mini_tool_parser",
+        "Phi4MiniJsonToolParser",
+    ),
+    "pythonic": (
+        "pythonic_tool_parser",
+        "PythonicToolParser",
+    ),
+    "qwen3_coder": (
+        "qwen3coder_tool_parser",
+        "Qwen3CoderToolParser",
+    ),
+    "qwen3_xml": (
+        "qwen3xml_tool_parser",
+        "Qwen3XMLToolParser",
+    ),
+    "seed_oss": (
+        "seed_oss_tool_parser",
+        "SeedOssToolParser",
+    ),
+    "step3": (
+        "step3_tool_parser",
+        "Step3ToolParser",
+    ),
+    "xlam": (
+        "xlam_tool_parser",
+        "xLAMToolParser",
+    ),
+    "gigachat3": (
+        "gigachat3_tool_parser",
+        "GigaChat3ToolParser",
+    ),
+}
+
+
+def register_lazy_tool_parsers():
+    for name, (file_name, class_name) in _TOOL_PARSERS_TO_REGISTER.items():
+        module_path = f"vllm.tool_parsers.{file_name}"
+        ToolParserManager.register_lazy_module(name, module_path, class_name)
+
+
+register_lazy_tool_parsers()

vllm/tool_parsers/abstract_tool_parser.py

@@ -17,12 +17,12 @@ from vllm.entrypoints.openai.protocol import (
     ResponsesRequest,
     ResponseTextConfig,
 )
-from vllm.entrypoints.openai.tool_parsers.utils import get_json_schema_from_tools
 from vllm.logger import init_logger
 from vllm.sampling_params import (
     StructuredOutputsParams,
 )
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.utils import get_json_schema_from_tools
 from vllm.utils.collection_utils import is_list_of
 from vllm.utils.import_utils import import_from_path
 
@@ -203,7 +203,7 @@ class ToolParserManager:
         Example:
             ToolParserManager.register_lazy_module(
                 name="kimi_k2",
-                module_path="vllm.entrypoints.openai.tool_parsers.kimi_k2_parser",
+                module_path="vllm.tool_parsers.kimi_k2_parser",
                 class_name="KimiK2ToolParser",
             )
         """

vllm/tool_parsers/deepseekv31_tool_parser.py

@@ -15,11 +15,9 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import ToolParser
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/deepseekv32_tool_parser.py

@@ -17,11 +17,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/deepseekv3_tool_parser.py

@@ -15,11 +15,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/ernie45_tool_parser.py

@@ -15,11 +15,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/gigachat3_tool_parser.py

@@ -16,9 +16,9 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import ToolParser
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import ToolParser
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/glm4_moe_tool_parser.py

@@ -18,11 +18,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/granite_20b_fc_tool_parser.py

@@ -19,17 +19,17 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.entrypoints.openai.tool_parsers.utils import (
+from vllm.tool_parsers.utils import (
     consume_space,
     find_common_prefix,
     is_complete_json,
     partial_json_loads,
 )
-from vllm.logger import init_logger
-from vllm.tokenizers import TokenizerLike
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/granite_tool_parser.py

@@ -17,17 +17,17 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.entrypoints.openai.tool_parsers.utils import (
+from vllm.tool_parsers.utils import (
     consume_space,
     find_common_prefix,
     is_complete_json,
     partial_json_loads,
 )
-from vllm.logger import init_logger
-from vllm.tokenizers import TokenizerLike
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/hermes_tool_parser.py

@@ -18,12 +18,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
 from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/hunyuan_a13b_tool_parser.py

@@ -17,12 +17,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
-from vllm.entrypoints.openai.tool_parsers.utils import consume_space
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
+from vllm.tool_parsers.utils import consume_space
 from vllm.utils import random_uuid
 
 logger = init_logger(__name__)

vllm/tool_parsers/internlm2_tool_parser.py

@@ -17,12 +17,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
-from vllm.entrypoints.openai.tool_parsers.utils import extract_intermediate_diff
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
+from vllm.tool_parsers.utils import extract_intermediate_diff
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/jamba_tool_parser.py

@@ -18,11 +18,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers import ToolParser
-from vllm.entrypoints.openai.tool_parsers.utils import extract_intermediate_diff
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
 from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers import ToolParser
+from vllm.tool_parsers.utils import extract_intermediate_diff
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/kimi_k2_tool_parser.py

@@ -15,11 +15,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/llama4_pythonic_tool_parser.py

@@ -18,10 +18,10 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.logger import init_logger
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/llama_tool_parser.py

@@ -20,15 +20,15 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.entrypoints.openai.tool_parsers.utils import (
+from vllm.tool_parsers.utils import (
     find_common_prefix,
     is_complete_json,
     partial_json_loads,
 )
-from vllm.logger import init_logger
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/longcat_tool_parser.py

@@ -3,8 +3,8 @@
 
 import regex as re
 
-from vllm.entrypoints.openai.tool_parsers.hermes_tool_parser import Hermes2ProToolParser
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.hermes_tool_parser import Hermes2ProToolParser
 
 
 class LongcatFlashToolParser(Hermes2ProToolParser):

vllm/tool_parsers/minimax_m2_tool_parser.py

@@ -17,11 +17,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/minimax_tool_parser.py

@@ -17,12 +17,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
-from vllm.entrypoints.openai.tool_parsers.utils import extract_intermediate_diff
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
+from vllm.tool_parsers.utils import extract_intermediate_diff
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/mistral_tool_parser.py

@@ -21,12 +21,12 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
 from vllm.tokenizers.mistral import MistralTokenizer
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/olmo3_tool_parser.py

@@ -18,10 +18,10 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.logger import init_logger
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/openai_tool_parser.py

@@ -12,10 +12,10 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.logger import init_logger
 
 if TYPE_CHECKING:
     from vllm.tokenizers import TokenizerLike

vllm/tool_parsers/phi4mini_tool_parser.py

@@ -16,10 +16,10 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.logger import init_logger
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/pythonic_tool_parser.py

@@ -19,10 +19,10 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.logger import init_logger
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
-from vllm.logger import init_logger
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/qwen3coder_tool_parser.py

@@ -18,11 +18,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/qwen3xml_tool_parser.py

@@ -19,11 +19,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/seed_oss_tool_parser.py

@@ -21,11 +21,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 
 logger = init_logger(__name__)
 

vllm/tool_parsers/step3_tool_parser.py

@@ -17,11 +17,11 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
-    ToolParser,
-)
 from vllm.logger import init_logger
 from vllm.tokenizers import TokenizerLike
+from vllm.tool_parsers.abstract_tool_parser import (
+    ToolParser,
+)
 from vllm.utils import random_uuid
 
 logger = init_logger(__name__)

vllm/tool_parsers/xlam_tool_parser.py

@@ -17,7 +17,7 @@ from vllm.entrypoints.openai.protocol import (
     FunctionCall,
     ToolCall,
 )
-from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
+from vllm.tool_parsers.abstract_tool_parser import (
     ToolParser,
 )
 from vllm.logger import init_logger