Implementation:Bentoml BentoML Tag Class

Knowledge Sources	Bentoml_BentoML
Domains	Core Framework, Versioning, Artifact Management
Last Updated	2026-02-13 15:00 GMT

Overview

Implements the Tag class used throughout BentoML to represent versioned identifiers for bentos and models in the format name:version.

Description

The Tag class is an attrs-based dataclass with name and optional version fields. Both fields are automatically lowercased and validated against a strict pattern: must be at most 63 characters, contain only lowercase alphanumeric characters, underscores, dashes, or dots, and start and end with an alphanumeric character. The class provides comparison operators (__eq__, __lt__, __hash__), string conversion (__str__ produces "name:version"), and several factory methods: from_str (parse "name:version" strings), from_taglike (accept either a Tag or string), and make_new_version (generate a UUID-based version using base32 encoding). The path method returns a POSIX path representation (name/version), and latest_path returns the path to the "latest" version file. The module also includes to_snake_case for converting CamelCase to snake_case, and registers cattrs structure/unstructure hooks for Tag serialization.

Usage

Use the Tag class anywhere you need to reference a BentoML artifact by name and version. It is the fundamental identifier type used by Store, Bento, Model, and related BentoML subsystems.

Code Reference

Source Location

• Repository: Bentoml_BentoML
• File: src/bentoml/_internal/tag.py
• Lines: 1-158

Signature

def to_snake_case(name: str) -> str: ...
def validate_tag_str(value: str): ...

@attr.define(slots=True)
class Tag:
    name: str
    version: t.Optional[str]

    def __init__(self, name: str, version: t.Optional[str] = None): ...
    def __str__(self): ...
    def __repr__(self): ...
    def __eq__(self, other: object) -> bool: ...
    def __lt__(self, other: object) -> bool: ...
    def __hash__(self) -> int: ...
    @classmethod
    def from_taglike(cls, taglike: t.Union["Tag", str]) -> "Tag": ...
    @classmethod
    def from_str(cls, tag_str: str) -> "Tag": ...
    def make_new_version(self) -> "Tag": ...
    def path(self) -> str: ...
    def latest_path(self) -> str: ...

Import

from bentoml._internal.tag import Tag
# or via the public API:
from bentoml import Tag

I/O Contract

Inputs

Name	Type	Required	Description
name	str	Yes	The artifact name; will be lowercased and validated
version	str or None	No	The version string; will be lowercased and validated if provided
tag_str	str	Yes (from_str)	String in "name:version" or "name" format to parse
taglike	Tag or str	Yes (from_taglike)	Either a Tag instance or a string to convert

Outputs

Name	Type	Description
Tag	Tag	A validated Tag instance with name and optional version
str	str	String representation as "name:version" or "name" (via __str__)
str	str	POSIX path as "name/version" (via path())

Usage Examples

from bentoml import Tag

# Create a tag
tag = Tag("my_model", "v1.0.0")
print(str(tag))  # "my_model:v1.0.0"

# Parse from string
tag = Tag.from_str("iris_classifier:abc123")
print(tag.name)     # "iris_classifier"
print(tag.version)  # "abc123"

# Generate a new version
base_tag = Tag("my_model")
versioned_tag = base_tag.make_new_version()
print(versioned_tag)  # "my_model:<auto-generated-version>"

# Get filesystem path
print(tag.path())         # "iris_classifier/abc123"
print(tag.latest_path())  # "iris_classifier/latest"

Related Pages

• Implementation:Bentoml_BentoML_Store_Base