scikitplot.mlflow#

MLflow UI | SERVER share the exact same settings.

Adds a project-level configuration mechanism so multiple scripts (e.g., train.py, hpo.py, predict.py) share the exact same MLflow settings, regardless of current working directory.

Examples

Quiskstart Template: Beginner workflow demo

>>> import os
>>> import scikitplot as sp
>>>
>>> # print(sp.mlflow.DEFAULT_PROJECT_MARKERS)
>>> # Walk upward from `start` until a directory containing any marker is found.
>>> # export SCIKITPLOT_PROJECT_MARKERS='[".git","pyproject.toml","README.txt","configs/mlflow.toml"]'
>>> os.environ["SCIKITPLOT_PROJECT_MARKERS"] = (
...     '[".git","pyproject.toml","README.txt","configs/mlflow.toml"]'
... )
>>> sp.mlflow.workflow(
...     profile="local",
...     open_ui_seconds=30,
...     experiment_name="my-first-project",
...     fmt="toml",
...     overwrite=True,  # If config already exists: ./configs/mlflow.toml (use overwrite=True).
... )

CLI

>>> # Walk upward from `start` until a directory containing any marker is found.
>>> # export SCIKITPLOT_PROJECT_MARKERS='[".git","pyproject.toml","README.txt","configs/mlflow.toml"]'
>>> python -m scikitplot.mlflow --profile local --open-ui-seconds 5

User guide. See the MLflow Workflow Automation section for further details.

Cli Helper#

`DEFAULT_MLFLOW_SERVER_FLAGS`	frozenset() -> empty frozenset object frozenset(iterable) -> frozenset object
`MlflowServerCliCaps`	Parsed capability set for `mlflow server` CLI flags.
`ensure_flags_supported`	Validate that all long-form CLI flags in `args` are supported.
`get_mlflow_server_cli_caps`	Get supported `mlflow server` flags for the installed MLflow version.

Compat Helper#

`import_mlflow`	Import MLflow lazily with a user-friendly error.
`resolve_download_artifacts`	Resolve a canonical artifact download function across MLflow versions.

Config Helper#

_config.

User guide. See the MLflow Workflow Automation section for further details.

`ServerConfig`	Configuration that maps directly to `mlflow server` CLI flags.
`SessionConfig`	Session-level configuration for `scikitplot.mlflow.session`.

Container Helper#

running_in_docker

Detect whether the current process is running in a Docker container.

Customization Helper#

`MlflowProvider`	A customizable provider for MLflow-like libraries.
`get_provider`	Retrieve the currently active MLflow provider.
`set_provider`	Set the active MLflow provider globally.
`use_provider`	Temporarily set the MLflow provider for a context block.

Env Helper#

`EnvSnapshot`	Full snapshot of process environment for strict restoration.
`apply_env`	Apply `.env` and explicit overrides to `os.environ`.
`parse_dotenv`	Parse a minimal `.env` file containing KEY=VALUE assignments.

Error Helper#

`MlflowCliIncompatibleError`	Raised when a requested `mlflow server` option is not supported by the installed MLflow.
`MlflowIntegrationError`	Base exception for scikitplot.mlflow errors.
`MlflowNotInstalledError`	Raised when MLflow is required but not installed.
`MlflowServerStartError`	Raised when the managed MLflow server fails to start or exits prematurely.
`SecurityPolicyViolationError`	Raised when an operation is rejected by the active `SecurityPolicy`.

Facade Helper#

`ArtifactsFacade`	Artifact helper facade bound to a specific MLflow client/URI.
`ModelsFacade`	Model helper facade bound to a session-bound MLflow client.

Project Helper#

Project configuration helpers for mlflow.

This module provides two distinct, deterministic responsibilities:

Project root discovery via marker files/directories (e.g., pyproject.toml or .git).
Project-level MLflow config I/O (TOML/YAML) that normalizes local paths so that multiple scripts (train/hpo/predict) behave consistently regardless of current working directory.

Notes

This module intentionally exposes a small public surface for marker customization, while keeping the underlying mutable default private. Users should prefer: - get_project_markers - set_project_markers - project_markers (context manager) - Environment override via SCIKITPLOT_PROJECT_MARKERS (strict JSON list of strings) - Config override via a TOML file containing [project].markers = [...]

Examples

Option A — temporary (best for automation pipelines)

>>> from scikitplot.mlflow._project import project_markers, find_project_root
>>>
>>> with project_markers(["pyproject.toml", ".git", "configs/mlflow.toml"]):
...     root = find_project_root()

Option B — environment (best for CI)

>>> import os
>>>
>>> # Default marker file-folder for auto detection
>>> # Walk upward from `start` until a directory containing any marker is found.
>>> # export SCIKITPLOT_PROJECT_MARKERS='[".git","pyproject.toml","README.txt","configs/mlflow.toml"]'
>>> os.environ["SCIKITPLOT_PROJECT_MARKERS"] = (
...     '[".git","pyproject.toml","README.txt","configs/mlflow.toml"]'
... )

Option C — config-driven (best for teams)

>>> [project]
>>> markers = ["pyproject.toml", ".git", "configs/mlflow.toml"]

User guide. See the MLflow Workflow Automation section for further details.

`DEFAULT_PROJECT_MARKERS`	Built-in immutable sequence.
`ProjectConfig`	Project-level configuration for MLflow usage across multiple scripts.
`dump_project_config_yaml`	Write a ProjectConfig to a YAML file.
`ensure_local_store_layout`	Ensure local backend/artifact directories exist.
`find_project_root`	Find a project root directory deterministically.
`get_project_markers`	Resolve project markers deterministically.
`load_project_config`	Load project MLflow config from TOML or YAML based on file extension.
`load_project_config_toml`	Load project MLflow config from a TOML file.
`load_project_config_yaml`	Load project MLflow config from a YAML file.
`normalize_mlflow_store_values`	Normalize local store values for consistent multi-script usage.
`project_markers`	Temporarily override module default markers for a block.
`set_project_markers`	Set the module default markers.

Readiness Helper#

wait_tracking_ready

Wait until the MLflow tracking REST API responds.

Security Helper#

`DEFAULT_SECURITY_POLICY`	Declarative security policy for `scikitplot.mlflow` operations.
`RELAXED_SECURITY_POLICY`	Declarative security policy for `scikitplot.mlflow` operations.
`SecurityPolicy`	Declarative security policy for `scikitplot.mlflow` operations.
`get_security_policy`	Return the currently active `SecurityPolicy`, or `None`.
`security_policy`	Temporarily activate a `SecurityPolicy` for a context block.
`set_security_policy`	Set the active `SecurityPolicy` globally.

Server Helper#

`SpawnedServer`	Spawned MLflow server process state.
`build_server_args`	Build the CLI args for `mlflow server` from a ServerConfig.
`build_server_command`	Build a deterministic `mlflow server` command.
`spawn_server`	Spawn an MLflow server subprocess.

Session Helper#

`MlflowHandle`	A handle that proxies the upstream `mlflow` module while adding session context.
`session`	Create a strict, context-managed MLflow session.
`session_from_file`	Create an MLflow session using a shared project config file (TOML or YAML).
`session_from_toml`	Create an MLflow session using a shared project TOML config.

Helper#

`MlflowVersion`	Parsed MLflow version.
`is_mlflow_installed`	Check whether MLflow is installed in the current Python environment.
`mlflow_version`	Retrieve the installed MLflow version (if available).

Workflow Helper#

`WorkflowPaths`	Standardized project config paths used by the workflow.
`builtin_config_path`	Return the path to the built-in demo config shipped with the package.
`default_project_paths`	Compute standard config file paths for a project.
`export_builtin_config`	Export the built-in demo config into the current project.
`patch_experiment_name_in_toml`	Patch `experiment_name = ...` in a TOML config deterministically.
`run_demo`	Run a beginner-friendly end-to-end demo workflow.
`workflow`	Run the built-in end-to-end MLflow workflow demo.