Tool Calling¶

vllm-mlx 支持与 OpenAI 兼容的 tool calling（function calling），并为多种主流模型系列提供自动解析。

快速开始¶

启动服务器时添加 --enable-auto-tool-choice 标志即可启用 tool calling：

vllm-mlx serve mlx-community/Devstral-Small-2507-4bit \
  --enable-auto-tool-choice \
  --tool-call-parser mistral

然后通过标准 OpenAI API 使用工具：

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")

response = client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "What's the weather in Paris?"}],
    tools=[{
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "Get weather for a city",
            "parameters": {
                "type": "object",
                "properties": {
                    "city": {"type": "string", "description": "City name"}
                },
                "required": ["city"]
            }
        }
    }]
)

# Check for tool calls
if response.choices[0].message.tool_calls:
    for tc in response.choices[0].message.tool_calls:
        print(f"Function: {tc.function.name}")
        print(f"Arguments: {tc.function.arguments}")

支持的 tool parser¶

使用 --tool-call-parser 为您的模型系列选择对应的 tool parser：

Parser	别名	模型	格式
`auto`		任意模型	自动检测格式（依次尝试所有 parser）
`mistral`		Mistral、Devstral	`[TOOL_CALLS]` JSON 数组
`qwen`	`qwen3`	Qwen、Qwen3	`<tool_call>` XML 或 `[Calling tool:]`
`llama`	`llama3`、`llama4`	Llama 3.x、4.x	`<function=name>` 标签
`hermes`	`nous`	Hermes、NousResearch	`<tool_call>` XML 包裹的 JSON
`deepseek`	`deepseek_v3`、`deepseek_r1`	DeepSeek V3、R1	Unicode 分隔符
`kimi`	`kimi_k2`、`moonshot`	Kimi K2、Moonshot	`<\\|tool_call_begin\\|>` 标记
`granite`	`granite3`	IBM Granite 3.x、4.x	`<\\|tool_call\\|>` 或 `<tool_call>`
`nemotron`	`nemotron3`	NVIDIA Nemotron	`<tool_call><function=...><parameter=...>`
`xlam`		Salesforce xLAM	含 `tool_calls` 数组的 JSON
`functionary`	`meetkai`	MeetKai Functionary	多个 function 块
`glm47`	`glm4`	GLM-4.7、GLM-4.7-Flash	`<tool_call>` 配合 `<arg_key>`/`<arg_value>` XML

模型示例¶

Mistral / Devstral¶

# Devstral Small（针对编程和 tool use 优化）
vllm-mlx serve mlx-community/Devstral-Small-2507-4bit \
  --enable-auto-tool-choice --tool-call-parser mistral

# Mistral Instruct
vllm-mlx serve mlx-community/Mistral-7B-Instruct-v0.3-4bit \
  --enable-auto-tool-choice --tool-call-parser mistral

Qwen¶

# Qwen3
vllm-mlx serve mlx-community/Qwen3-4B-4bit \
  --enable-auto-tool-choice --tool-call-parser qwen

Llama¶

# Llama 3.2
vllm-mlx serve mlx-community/Llama-3.2-3B-Instruct-4bit \
  --enable-auto-tool-choice --tool-call-parser llama

DeepSeek¶

# DeepSeek V3
vllm-mlx serve mlx-community/DeepSeek-V3-0324-4bit \
  --enable-auto-tool-choice --tool-call-parser deepseek

IBM Granite¶

# Granite 4.0
vllm-mlx serve mlx-community/granite-4.0-tiny-preview-4bit \
  --enable-auto-tool-choice --tool-call-parser granite

NVIDIA Nemotron¶

# Nemotron 3 Nano
vllm-mlx serve mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-6Bit \
  --enable-auto-tool-choice --tool-call-parser nemotron

GLM-4.7¶

# GLM-4.7 Flash
vllm-mlx serve lmstudio-community/GLM-4.7-Flash-MLX-8bit \
  --enable-auto-tool-choice --tool-call-parser glm47

Kimi K2¶

# Kimi K2
vllm-mlx serve mlx-community/Kimi-K2-Instruct-4bit \
  --enable-auto-tool-choice --tool-call-parser kimi

Salesforce xLAM¶

# xLAM
vllm-mlx serve mlx-community/xLAM-2-fc-r-4bit \
  --enable-auto-tool-choice --tool-call-parser xlam

Auto Parser¶

如果不确定使用哪个 tool parser，auto parser 会尝试自动检测格式：

vllm-mlx serve mlx-community/Qwen3-4B-4bit \
  --enable-auto-tool-choice --tool-call-parser auto

auto parser 按以下顺序依次尝试各种格式：

Mistral（[TOOL_CALLS]）
Qwen 括号格式（[Calling tool:]）
Nemotron（<tool_call><function=...><parameter=...>）
Qwen/Hermes XML（<tool_call>{...}</tool_call>）
Llama（<function=name>{...}</function>）
原始 JSON

Streaming Tool Calls¶

Tool calling 支持 streaming。模型生成完毕后发送 tool call 信息：

stream = client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "What's 25 * 17?"}],
    tools=[{
        "type": "function",
        "function": {
            "name": "calculator",
            "description": "Calculate math expressions",
            "parameters": {
                "type": "object",
                "properties": {
                    "expression": {"type": "string"}
                },
                "required": ["expression"]
            }
        }
    }],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.tool_calls:
        for tc in chunk.choices[0].delta.tool_calls:
            print(f"Tool call: {tc.function.name}({tc.function.arguments})")

处理工具返回结果¶

收到 tool call 后，执行对应函数并将结果返回给模型：

import json

# 第一次请求，模型决定调用工具
response = client.chat.completions.create(
    model="default",
    messages=[{"role": "user", "content": "What's the weather in Tokyo?"}],
    tools=[weather_tool]
)

# 获取 tool call
tool_call = response.choices[0].message.tool_calls[0]
tool_call_id = tool_call.id
function_name = tool_call.function.name
arguments = json.loads(tool_call.function.arguments)

# 执行函数（由您自行实现）
result = get_weather(**arguments)  # {"temperature": 22, "condition": "sunny"}

# 将结果返回给模型
response = client.chat.completions.create(
    model="default",
    messages=[
        {"role": "user", "content": "What's the weather in Tokyo?"},
        {"role": "assistant", "tool_calls": [tool_call]},
        {"role": "tool", "tool_call_id": tool_call_id, "content": json.dumps(result)}
    ],
    tools=[weather_tool]
)

print(response.choices[0].message.content)
# "The weather in Tokyo is sunny with a temperature of 22C."

Think 标签处理¶

会产生 <think>...</think> reasoning 标签的模型（如 DeepSeek-R1、Qwen3、GLM-4.7）均可自动处理。tool parser 会在提取 tool call 前剥离 thinking 内容，因此 reasoning 标签不会干扰 tool call 解析。

即使 <think> 是通过提示词注入的（即仅有闭合标签 </think> 的隐式 think 标签），也同样适用。

CLI 参数参考¶

选项	说明
`--enable-auto-tool-choice`	启用自动 tool calling
`--tool-call-parser`	选择 tool parser（见上表）

完整选项请参阅 CLI Reference。