Advanced Topics of the Structural Tag

Built-in Structural Tag: get_builtin_structural_tag

get_builtin_structural_tag generates a StructuralTag for the given model type with the specified tools and options. The returned structural tag can be used with Grammar.from_structural_tag or GrammarCompiler.compile_structural_tag to obtain a grammar that matches the function-calling format in the corresponding model’s style within any-form text.

Use it when you need to constrain the model to output in a fixed pattern such as “tool name + parameter JSON”, e.g. for Llama, Qwen, Kimi, DeepSeek, or OpenAI Harmony, etc.

Parameters

  • model (BuiltinModels): The model type. Supported values:

    • "llama": Llama-style (e.g. Llama 3, Llama 4)

    • "qwen": Qwen-style (e.g. Qwen3)

    • "qwen_coder": Qwen Coder-style (e.g. Qwen3-Coder, Qwen3-Coder-Next)

    • "kimi": Kimi-style (e.g. Kimi-K2, Kimi-K2.5)

    • "deepseek_r1": DeepSeek-style (e.g. DeepSeek-V3.1, DeepSeek-R1, DeepSeek-V3.2-exp)

    • "harmony": OpenAI Harmony Response Format (e.g. gpt-oss)

    • "deepseek_v3_2": DeepSeek-V3.2-style (e.g. DeepSeek-V3.2)

    • "minimax": MiniMax-style (e.g. MiniMax-M2)

    • "glm47": GLM-style (e.g. GLM-5, GLM-4.7)

  • reasoning (bool, optional): Whether to enable reasoning mode (<think>/</think> tags). Default True.

  • tools (List[Dict[str, Any]], optional): List of tools; each item is a dict with a "function" key. The "function" dict must contain a "name" (string), and may contain:

    • "parameters": JSON Schema, which can be:

      • a dict (regular JSON Schema object), for example {"type": "object", "properties": {...}}

      • a bool: True means “any JSON value is accepted”, False means “no value is accepted”. If "parameters" is missing, the no constraint will be applied.

    • "strict": bool. Controls whether the parameter constraints are applied. When False, only the function name will be enforced, but the parameters is unconstrained. Default: True. Default value is [].

  • builtin_tools (List[Dict[str, Any]], optional): List of built-in tools (used only for "harmony"); each element has the same structure as items in tools. Default [].

  • force_empty_reasoning (bool, optional): When reasoning is on, whether to force empty thinking content at the beginning. Default False.

Passing an unsupported model raises ValueError.

Returns

StructuralTag: The structural tag for the given model’s function-calling format.

Example

from xgrammar import Grammar, get_builtin_structural_tag

tools = [
    {"function": {"name": "get_weather", "parameters": {"type": "object", "properties": {"city": {"type": "string"}}}}},
    {"function": {"name": "get_time", "parameters": {"type": "object", "properties": {}}}},
]

# Get the Llama-style structural tag and build a grammar
structural_tag = get_builtin_structural_tag("llama", tools=tools)
grammar = Grammar.from_structural_tag(structural_tag)

For the Harmony format you must provide both tools and builtin_tools:

structural_tag = get_builtin_structural_tag(
    "harmony",
    tools=[
        # User tool in strict mode, with a full JSON Schema
        {
            "function": {
                "name": "user_tool",
                "parameters": {"type": "object", "properties": {"q": {"type": "string"}}},
            }
        },
        # User tool in non-strict mode: name only, arguments unconstrained (equivalent to parameters=True)
        {"function": {"name": "user_tool_untyped", "strict": False}},
    ],
    builtin_tools=[
        # Built-in tools support the same strict / parameters combinations
        {
            "function": {
                "name": "builtin_tool",
                "parameters": {"type": "object", "properties": {}},
            }
        },
    ],
)
grammar = Grammar.from_structural_tag(structural_tag)

For formats that support reasoning (like Qwen3, Deepseek-R1, Kimi-k2-thinking ), pass reasoning to enable/disable the reasoning mode:

structural_tag = get_builtin_structural_tag("qwen", tools=tools, reasoning=True)
grammar = Grammar.from_structural_tag(structural_tag)

If reasoning is not passed, reasoning mode is enabled by default. Besides, when reasoning is True, you can also set force_empty_reasoning to constrain the reasoning content.

Supported models: get_builtin_structural_tag_supported_models

get_builtin_structural_tag_supported_models returns the supported model list for each built-in structural tag function. Call it with no args to get Dict[str, List[str]] (style → models), or pass a style name (e.g. "llama", "qwen") to get List[str] for that style. Use it to confirm which style a model uses before calling get_builtin_structural_tag.

Automatic End Detection

Certain “unlimited” formats — any_text, any_tokens, triggered_tags, and token_triggered_tags — can consume an unbounded amount of output. To know when to stop, the structural tag compiler automatically detects the end condition of the enclosing tag and adds it to the format’s internal exclude set.

The detection works by walking up to the nearest enclosing tag:

  • If the tag’s end is a string (or list of strings), the end strings are added to the exclude set of string-level unlimited formats (any_text, triggered_tags).

  • If the tag’s end is a token format, the end token ID is added to the exclude set of token-level unlimited formats (any_tokens, token_triggered_tags, exclude_token).

Currently only string–string and token–token pairs are detected. Cross-level detection (e.g. a token end for a string-level format) is not supported. If you need additional exclusions beyond what automatic detection provides, specify them explicitly via the format’s own exclude field (excludes for any_text, exclude_tokens for exclude_token/any_tokens/token_triggered_tags).

Deprecated API: Grammar.from_structural_tag(tags, triggers)

The deprecated API is still available for backward compatibility. However, it is recommended to use the new API instead.

Create a grammar from structural tags. The structural tag handles the dispatching of different grammars based on the tags and triggers: it initially allows any output, until a trigger is encountered, then dispatch to the corresponding tag; when the end tag is encountered, the grammar will allow any following output, until the next trigger is encountered.

The tags parameter is used to specify the output pattern. It is especially useful for LLM function calling, where the pattern is: <function=func_name>{"arg1": ..., "arg2": ...}</function>. This pattern consists of three parts: a begin tag (<function=func_name>), a parameter list according to some schema ({"arg1": ..., "arg2": ...}), and an end tag (</function>). This pattern can be described in a StructuralTagItem with a begin tag, a schema, and an end tag. The structural tag is able to handle multiple such patterns by passing them into multiple tags.

The triggers parameter is used to trigger the dispatching of different grammars. The trigger should be a prefix of a provided begin tag. When the trigger is encountered, the corresponding tag should be used to constrain the following output. There can be multiple tags matching the same trigger. Then if the trigger is encountered, the following output should match one of the tags. For example, in function calling, the triggers can be ["<function="]. Then if "<function=" is encountered, the following output must match one of the tags (e.g. <function=get_weather>{"city": "Beijing"}</function>).

The correspondence of tags and triggers is automatically determined: all tags with the same trigger will be grouped together. User should make sure any trigger is not a prefix of another trigger: then the correspondence of tags and triggers will be ambiguous.

To use this grammar in grammar-guided generation, the GrammarMatcher constructed from structural tag will generate a mask for each token. When the trigger is not encountered, the mask will likely be all-1 and not have to be used (fill_next_token_bitmask returns False, meaning no token is masked). When a trigger is encountered, the mask should be enforced (fill_next_token_bitmask will return True, meaning some token is masked) to the output logits.

The benefit of this method is the token boundary between tags and triggers is automatically handled. The user does not need to worry about the token boundary.

Parameters (deprecated)

  • tags (List[StructuralTagItem]): The structural tags.

  • triggers (List[str]): The triggers.

Returns (deprecated)

  • grammar (Grammar): The constructed grammar.

Example (deprecated)

from pydantic import BaseModel
from typing import List
from xgrammar import Grammar, StructuralTagItem

class Schema1(BaseModel):
    arg1: str
    arg2: int

class Schema2(BaseModel):
    arg3: float
    arg4: List[str]

tags = [
    StructuralTagItem(begin="<function=f>", schema=Schema1, end="</function>"),
    StructuralTagItem(begin="<function=g>", schema=Schema2, end="</function>"),
]
triggers = ["<function="]
grammar = Grammar.from_structural_tag(tags, triggers)