Implementation:Mlflow Mlflow Pyproject Generator

Knowledge Sources	Mlflow_Mlflow
Domains	Build, Packaging
Last Updated	2026-02-13 20:00 GMT

Overview

Generates all pyproject.toml files for MLflow's package variants (dev, release, skinny, tracing) from YAML requirement specifications, serving as the single source of truth for package configurations.

Description

This script constructs and writes pyproject.toml files for four distinct MLflow package variants, each defined by the PackageType enum:

• SKINNY -- The lightweight mlflow-skinny package without SQL storage, server, UI, or data science dependencies. Written to libs/skinny/pyproject.toml with a generated README.
• TRACING -- The mlflow-tracing package containing only the minimum set of modules for instrumenting code with MLflow Tracing. Written to libs/tracing/pyproject.toml with a curated include/exclude file list (TRACING_INCLUDE_FILES / TRACING_EXCLUDE_FILES).
• RELEASE -- The full mlflow release package that depends on mlflow-skinny and mlflow-tracing at the same version, plus core requirements. Written to pyproject.release.toml.
• DEV -- The development mlflow package without sub-package dependencies (skinny and tracing requirements are included directly). Written to pyproject.toml, preserving the manual dev tool settings section below a separator.

The script reads dependencies from YAML files in the requirements/ directory using pydantic for validation via the PackageRequirement model. It validates that tracing requirements are a strict subset of skinny requirements. Each requirement spec includes pip_release name, max_major_version, optional minimum version, unsupported versions, environment markers, and extras.

Output is formatted with taplo (the TOML formatter installed by bin/install.py) and written only if content has changed to avoid unnecessary diffs. The DEV variant also runs uv lock to update the lockfile.

Usage

Run this script whenever dependencies change in the YAML requirement files, or when the package version is bumped. It requires taplo to be installed (via python bin/install.py).

Code Reference

Source Location

• Repository: Mlflow_Mlflow
• File: dev/pyproject.py
• Lines: 1-506

Signature

class PackageType(Enum):
    SKINNY = "skinny"
    RELEASE = "release"
    DEV = "dev"
    TRACING = "tracing"
    def description(self) -> str: ...

class PackageRequirement(BaseModel):
    pip_release: str
    max_major_version: int
    minimum: str | None = None
    unsupported: list[str] | None = None
    markers: str | None = None
    extras: list[str] | None = None
    freeze: bool | None = None

def generate_requirements_from_yaml(requirements_yaml: RequirementsYaml) -> list[str]: ...
def read_requirements_yaml(yaml_path: Path) -> list[str]: ...
def build(package_type: PackageType) -> None: ...
def write_toml_file_if_changed(file_path: Path, description: str, toml_data: dict[str, Any]) -> None: ...
def format_content_with_taplo(content: str) -> str: ...
def main() -> None: ...

Import

# Run from repository root (requires taplo to be installed)
python dev/pyproject.py

I/O Contract

Inputs

Name	Type	Required	Description
requirements/skinny-requirements.yaml	YAML file	Yes	Dependencies for the mlflow-skinny package
requirements/tracing-requirements.yaml	YAML file	Yes	Dependencies for the mlflow-tracing package
requirements/core-requirements.yaml	YAML file	Yes	Core dependencies for the full mlflow package
requirements/gateway-requirements.yaml	YAML file	Yes	Dependencies for the gateway optional extra
requirements/genai-requirements.yaml	YAML file	Yes	Dependencies for the genai optional extra
mlflow/version.py	Python file	Yes	Contains the VERSION string for the package
.python-version	Text file	Yes	Minimum Python version requirement
mlflow/ml-package-versions.yml	YAML file	Yes	Supported ML framework version ranges

Outputs

Name	Type	Description
pyproject.toml	TOML file	Development package configuration (preserves manual dev tool settings)
pyproject.release.toml	TOML file	Release package configuration with mlflow-skinny and mlflow-tracing dependencies
libs/skinny/pyproject.toml	TOML file	mlflow-skinny package configuration
libs/tracing/pyproject.toml	TOML file	mlflow-tracing package configuration
libs/skinny/README_SKINNY.md	Markdown file	Generated README for the skinny package
uv.lock	Lock file	Updated dependency lock file (for DEV variant only)

Usage Examples

Basic Usage

# Generate all pyproject.toml variants
python dev/pyproject.py

# Ensure taplo is installed first
python bin/install.py taplo
python dev/pyproject.py

Related Pages

• Environment:Mlflow_Mlflow_Python_Runtime_Environment