SessionConfig#

class scikitplot.mlflow.SessionConfig(tracking_uri=None, public_tracking_uri=None, registry_uri=None, env_file=None, extra_env=None, startup_timeout_s=30.0, ensure_reachable=False, experiment_name=None, create_experiment_if_missing=True, default_run_name=None, default_run_tags=None)[source]#

Session-level configuration for scikitplot.mlflow.session.

Parameters:
tracking_uristr or None, default=None

MLflow tracking URI to use inside the session.

public_tracking_uristr or None, default=None

Optional UI URL to display to users (e.g., when running in Docker/remote notebooks). This does NOT change the internal tracking URI used by MLflow calls; it is only for human-facing navigation. If None, the session reads MLFLOW_TRACKING_URI from the environment (after loading env_file, if provided). If start_server=True and no URI is provided, it will be constructed as http://{tracking_host}:{server.port}.

registry_uristr or None, default=None

Optional MLflow registry URI. If None, the session reads MLFLOW_REGISTRY_URI from the environment (after loading env_file, if provided).

env_filestr or None, default=None

Optional path to a .env file. Keys are loaded only if missing from os.environ, matching MLflow CLI --env-file behavior.

extra_envMapping[str, str] or None, default=None

Additional environment variables to set for the duration of the session. These override current env vars and are restored on exit.

startup_timeout_sfloat, default=30.0

Maximum seconds to wait for MLflow server readiness when start_server=True.

ensure_reachablebool, default=False

If True, verify the configured tracking URI is reachable even when start_server=False. This performs the same readiness check used for managed servers.

experiment_namestr or None, default=None

If provided, set the active experiment on session entry (before any runs).

create_experiment_if_missingbool, default=True

Controls behavior when experiment_name does not exist: - True: create the experiment (via mlflow.set_experiment) - False: raise a KeyError (strict fail)

default_run_namestr or None, default=None

Default run_name applied by MlflowHandle.start_run if the caller did not provide an explicit run name.

default_run_tagsMapping[str, str] or None, default=None

Default tags applied by MlflowHandle.start_run when a run begins.

Raises:
ValueError

If startup_timeout_s is not positive.

Parameters:

  • tracking_uri (str | None)

  • public_tracking_uri (str | None)

  • registry_uri (str | None)

  • env_file (str | None)

  • extra_env (Mapping[str, str] | None)

  • startup_timeout_s (float)

  • ensure_reachable (bool)

  • experiment_name (str | None)

  • create_experiment_if_missing (bool)

  • default_run_name (str | None)

  • default_run_tags (Mapping[str, str] | None)

Notes

Precedence order is strict: 1) explicit arguments (tracking_uri, registry_uri, extra_env) 2) existing environment variables 3) .env file (fills missing keys only) 4) defaults

create_experiment_if_missing: bool = True#
default_run_name: str | None = None#
default_run_tags: Mapping[str, str] | None = None#
ensure_reachable: bool = False#
env_file: str | None = None#
experiment_name: str | None = None#
extra_env: Mapping[str, str] | None = None#
public_tracking_uri: str | None = None#
registry_uri: str | None = None#
startup_timeout_s: float = 30.0#
tracking_uri: str | None = None#