From 0a9bc666442a7010c42cedc7707773636bbbb8d2 Mon Sep 17 00:00:00 2001
From: bartzbeielstein <32470350+bartzbeielstein@users.noreply.github.com>
Date: Sat, 13 Jun 2026 01:03:25 +0200
Subject: [PATCH] feat(tasks)!: remove the tasks subpackage; relocate ENTSO-E
 CLI to spotforecast2.entsoe_cli

The last tasks module, task_entsoe, is dissolved:

- entsoe_data_loader / entsoe_test_data_loader now live in
  spotforecast2_safe.data.entsoe_loader (safe 22.3.0); the pin floor
  rises to >=22.3.0.
- entsoe_lgbm_factory was body-identical to
  spotforecast2_safe.multitask.factories.default_lgbm_forecaster_factory
  and is dropped in favour of the safe original.
- The CLI (main, entsoe_xgb_factory, pipeline plumbing) moves to
  spotforecast2.entsoe_cli; the spotforecast2-entsoe console script is
  unchanged in name and behaviour.
- bart26k-lecture migrated its 8 import sites ahead of this change
  (commit 1df90d3 there); the team4 consumer-contract gate passes 4/4
  against the updated qmd.
- docs/tasks/entsoe.qmd loses two sections that documented the legacy
  ForecasterRecursiveLGBM/XGB wrapper API removed back in ADR-002; they
  are rewritten around the current factory + MultiTask pipeline.

BREAKING CHANGE: the spotforecast2.tasks subpackage is removed.
Import the data loaders from spotforecast2_safe.data.entsoe_loader, the
LightGBM factory from spotforecast2_safe.multitask.factories, and the
CLI module as spotforecast2.entsoe_cli. The spotforecast2-entsoe console
script is unaffected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 _quarto.yml                                   |  12 +-
 docs/multitask/entsoe.qmd                     |   4 +-
 docs/tasks/entsoe.qmd                         | 133 ++++++------
 pyproject.toml                                |  10 +-
 .../{tasks/task_entsoe.py => entsoe_cli.py}   | 192 ++----------------
 src/spotforecast2/tasks/__init__.py           |   2 -
 ..._cadence.py => test_entsoe_cli_cadence.py} |  30 +--
 ..._predict.py => test_entsoe_cli_predict.py} |  17 +-
 ...tsoe_train.py => test_entsoe_cli_train.py} |  21 +-
 tests/test_entsoe_pipeline.py                 |   8 +-
 uv.lock                                       |  10 +-
 11 files changed, 126 insertions(+), 313 deletions(-)
 rename src/spotforecast2/{tasks/task_entsoe.py => entsoe_cli.py} (57%)
 delete mode 100644 src/spotforecast2/tasks/__init__.py
 rename tests/{test_task_entsoe_cadence.py => test_entsoe_cli_cadence.py} (70%)
 rename tests/{test_tasks_entsoe_predict.py => test_entsoe_cli_predict.py} (76%)
 rename tests/{test_tasks_entsoe_train.py => test_entsoe_cli_train.py} (78%)

diff --git a/_quarto.yml b/_quarto.yml
index 69b46340..5af11cd4 100644
--- a/_quarto.yml
+++ b/_quarto.yml
@@ -130,10 +130,10 @@ website:
             contents:
               - text: "autocorrelation"
                 file: docs/reference/stats.autocorrelation.qmd
-          - section: "Tasks"
+          - section: "ENTSO-E CLI"
             contents:
-              - text: "task_entsoe"
-                file: docs/reference/tasks.task_entsoe.qmd
+              - text: "entsoe_cli"
+                file: docs/reference/entsoe_cli.qmd
 
       - section: "Processing Guides"
         contents:
@@ -251,10 +251,10 @@ quartodoc:
       contents:
         - stats.autocorrelation
 
-    - title: "Tasks"
-      desc: "ENTSO-E forecasting task and CLI."
+    - title: "ENTSO-E CLI"
+      desc: "ENTSO-E download / train / predict command-line pipeline."
       contents:
-        - tasks.task_entsoe
+        - entsoe_cli
 
     - title: "Warnings"
       desc: "Warning-style configuration for spotforecast2."
diff --git a/docs/multitask/entsoe.qmd b/docs/multitask/entsoe.qmd
index 396c3d50..91a26c43 100644
--- a/docs/multitask/entsoe.qmd
+++ b/docs/multitask/entsoe.qmd
@@ -30,9 +30,11 @@ import pandas as pd
 warnings.filterwarnings("ignore")
 
 from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.multitask.factories import (
+    default_lgbm_forecaster_factory as entsoe_lgbm_factory,
+)
 
 from spotforecast2.multitask.multi import MultiTask
-from spotforecast2.tasks.task_entsoe import entsoe_lgbm_factory
 
 CACHE_HOME = tempfile.mkdtemp()
 ```
diff --git a/docs/tasks/entsoe.qmd b/docs/tasks/entsoe.qmd
index 3e90372a..9b7bde4a 100644
--- a/docs/tasks/entsoe.qmd
+++ b/docs/tasks/entsoe.qmd
@@ -275,107 +275,94 @@ print(ts_clean.values)  # [1.0, 2.0, 3.0, 4.0, 5.0]
 
 ## Forecaster Models
 
+The pipeline builds its forecasters through *factory* functions that take a
+`ConfigEntsoe` and return a ready-to-fit
+`spotforecast2_safe.forecaster.recursive.ForecasterRecursive`.
+
 ### LightGBM Forecaster
 
-Create a LightGBM-based recursive forecaster:
+The stock LightGBM factory lives in the safe package:
 
-```python
-from spotforecast2.tasks.task_entsoe import ForecasterRecursiveLGBM, config
+```{python}
+from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
 
-model = ForecasterRecursiveLGBM(iteration=1)
+config = ConfigEntsoe()
+forecaster = default_lgbm_forecaster_factory(config)
 
-print(model.name)             # 'lgbm'
-print(model.random_state)     # 314159 (from config)
-print(len(model.preprocessor.periods))  # 5 (from config)
+print(type(forecaster).__name__)
+print("lags:", len(forecaster.lags))
 ```
 
 ### XGBoost Forecaster
 
-Create an XGBoost-based recursive forecaster:
-
-```python
-from spotforecast2.tasks.task_entsoe import ForecasterRecursiveXGB, config
-
-model = ForecasterRecursiveXGB(iteration=1, lags=24)
+The XGBoost variant ships with the CLI module (xgboost is not a
+safe-package dependency):
 
-print(model.name)  # 'xgb'
-```
-
-### Custom Configuration Forecaster
-
-Override default configuration values:
-
-```python
-from spotforecast2.tasks.task_entsoe import ForecasterRecursiveLGBM
-from spotforecast2_safe.data import Period
+```{python}
+from spotforecast2_safe.configurator import ConfigEntsoe
 
-custom_periods = [
-    Period(name='hourly', n_periods=24, column='hour', input_range=(1, 24)),
-]
+from spotforecast2.entsoe_cli import entsoe_xgb_factory
 
-model = ForecasterRecursiveLGBM(
-    iteration=1,
-    lags=48,
-    periods=custom_periods,
-    country_code='FR',
-    random_state=42
-)
+config = ConfigEntsoe()
+forecaster = entsoe_xgb_factory(config)
 
-print(len(model.preprocessor.periods))  # 1
-print(model.preprocessor.country_code)  # 'FR'
+print(type(forecaster.estimator).__name__)
 ```
 
----
+Both factories honour `config.random_state`, `config.lags_consider`, and
+`config.window_size`; supply your own factory through
+`config.forecaster_factory` to customise further (see the Multitask
+tutorial).
 
 ## Using the Python API (Notebooks & Quarto)
 
 ### Full Prediction Pipeline
 
-For users working in Jupyter Notebooks or Quarto, the entire ENTSO-E pipeline can be executed using the Python API. This approach is highly recommended for safety-critical research as it allows for precise control over time windows and hyperparameters.
+For users working in Jupyter Notebooks or Quarto, the entire ENTSO-E pipeline
+can be executed through `MultiTask` — the same path the CLI's `train` and
+`predict` subcommands take. This approach is recommended for research as it
+gives precise control over time windows and hyperparameters.
 
 ```python
-import pandas as pd
+import logging
 import os
+
+from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.data.entsoe_loader import entsoe_data_loader
+from spotforecast2_safe.data.fetch_data import get_cache_home
 from spotforecast2_safe.downloader.entsoe import download_new_data
-from spotforecast2_safe.manager.trainer import handle_training as handle_training_safe
-from spotforecast2_safe.manager.predictor import get_model_prediction as get_model_prediction_safe
-from spotforecast2.plots.plotter import make_plot
-from spotforecast2.tasks.task_entsoe import ForecasterRecursiveLGBM
-
-# 1. Setup Time Windows (Last 3 years until last month) and country:
-country_code = "ES"  
-now = pd.Timestamp.now(tz='UTC').floor('D')
-current_month_start = now.replace(day=1)
-last_month_start = (current_month_start - pd.Timedelta(days=1)).replace(day=1)
-
-# 2. Download Data (Optional, requires ENTSOE_API_KEY)
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
+
+from spotforecast2.multitask import MultiTask
+
+# 1. Download data (optional, requires ENTSOE_API_KEY)
 api_key = os.environ.get("ENTSOE_API_KEY")
 if api_key:
-    download_new_data(api_key=api_key, start="202301010000", country_code=country_code)
-
-# 3. Configure and Train
-# Explicit parameters override global configuration for reproducibility
-model_class = ForecasterRecursiveLGBM
-model_name = "lgbm_advanced"
-
-handle_training_safe(
-    model_class=model_class,
-    model_name=model_name,
-    train_size=pd.Timedelta(days=3 * 365),
-    end_dev=last_month_start.strftime("%Y-%m-%d %H:%M%z"),
-    country_code=country_code
-)
+    download_new_data(api_key=api_key, start="202301010000")
 
-# 4. Generate Predictions for the forecast horizon
-# The predictor will automatically load the model trained above
-predictions = get_model_prediction_safe(
-    model_name=model_name,
-    predict_size=24 * 31
+# 2. Wire the loader and factory into the config
+config = ConfigEntsoe()
+config.targets = ["Actual Load"]
+config.agg_weights = [1.0]
+config.bounds = [(-1e9, 1e9)]
+config.data_loader = entsoe_data_loader
+config.forecaster_factory = default_lgbm_forecaster_factory
+config.data_frame_name = "entsoe-lgbm"
+
+# 3. Run the five-step pipeline (task="defaults" trains; "predict" reuses
+#    the saved model)
+mt = MultiTask(
+    config,
+    task="defaults",
+    cache_home=get_cache_home(config.cache_home),
+    log_level=logging.ERROR,
 )
-
-# 5. Visualize Results
-if predictions:
-    make_plot(predictions)
+mt.prepare_data()
+mt.detect_outliers()
+mt.impute()
+mt.build_exogenous_features()
+mt.run(show=True)
 ```
 
 ---
diff --git a/pyproject.toml b/pyproject.toml
index ad25cd56..72c4ef2a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -34,16 +34,16 @@ dependencies = [
     # 22.0.0 made warm_start_lags the seed lag list itself (default
     # DEFAULT_WARM_START_LAGS, None disables) — consumed by SpotOptimStrategy.
     # 21.2.0 added max_time_spotoptim, forwarded as SpotOptim's max_time.
-    "spotforecast2-safe>=22.0.0,<23",
+    "spotforecast2-safe>=22.3.0,<23",
     # spotoptim 1.0 is sequential-only and lean: torch/tensorboard moved to its
     # ``[torch]`` extra. sf2 forwards tensorboard_* kwargs into SpotOptim, so we
     # pin the extra to keep the TensorBoard tuning dashboards working (they were
     # always available via spotoptim's old hard torch dependency).
     "spotoptim[torch]>=1.0.0,<2",
     "tqdm>=4.67.2",
-    # Directly imported by spotforecast2.tasks.task_entsoe (and the xgb
-    # forecaster model). Previously satisfied transitively; declared explicitly
-    # now that spotoptim 1.0 no longer pulls it in.
+    # Directly imported by spotforecast2.entsoe_cli (and the xgb forecaster
+    # model). Previously satisfied transitively; declared explicitly now that
+    # spotoptim 1.0 no longer pulls it in.
     "xgboost>=3.2.0",
 ]
 
@@ -74,7 +74,7 @@ dev = [
 ]
 
 [project.scripts]
-spotforecast2-entsoe = "spotforecast2.tasks.task_entsoe:main"
+spotforecast2-entsoe = "spotforecast2.entsoe_cli:main"
 
 [tool.uv]
 # Accept pre-releases ONLY for dependencies whose specifier carries an explicit
diff --git a/src/spotforecast2/tasks/task_entsoe.py b/src/spotforecast2/entsoe_cli.py
similarity index 57%
rename from src/spotforecast2/tasks/task_entsoe.py
rename to src/spotforecast2/entsoe_cli.py
index d989a3d2..fb938fa3 100644
--- a/src/spotforecast2/tasks/task_entsoe.py
+++ b/src/spotforecast2/entsoe_cli.py
@@ -5,7 +5,10 @@
 
 Drives the pipeline directly through `spotforecast2.multitask.multi.MultiTask`
 with `ConfigEntsoe` plugged in via the `data_loader` and `forecaster_factory`
-hooks introduced in ADR-001.  The CLI exposes four subcommands:
+hooks introduced in ADR-001.  The data loaders live in
+`spotforecast2_safe.data.entsoe_loader`; the LightGBM factory is the stock
+`spotforecast2_safe.multitask.factories.default_lgbm_forecaster_factory`.
+The CLI exposes four subcommands:
 
 - ``download`` — fetch raw data from the ENTSO-E Transparency Platform.
 - ``merge`` — concatenate raw CSVs into the interim merged file.
@@ -30,12 +33,13 @@
 from typing import Any, Optional
 
 import pandas as pd
-from lightgbm import LGBMRegressor
 from spotforecast2_safe.configurator import ConfigEntsoe
-from spotforecast2_safe.data.fetch_data import get_cache_home, get_data_home
+from spotforecast2_safe.data.entsoe_loader import entsoe_data_loader
+from spotforecast2_safe.data.fetch_data import get_cache_home
 from spotforecast2_safe.downloader.entsoe import download_new_data, merge_build_manual
 from spotforecast2_safe.forecaster.recursive import ForecasterRecursive
 from spotforecast2_safe.manager.trainer import should_retrain
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
 from spotforecast2_safe.preprocessing import RollingFeatures
 from xgboost import XGBRegressor
 
@@ -57,176 +61,6 @@
 _DEFAULT_AGG_WEIGHTS = [1.0]
 
 
-def entsoe_data_loader(config: ConfigEntsoe) -> pd.DataFrame:
-    """Read the merged interim ENTSO-E CSV that ``config.data_filename`` points at.
-
-    Args:
-        config: A `ConfigEntsoe` with ``data_filename`` set.  Relative paths
-            are resolved against `spotforecast2_safe.data.fetch_data.get_data_home`.
-
-    Returns:
-        DataFrame indexed by the ENTSO-E timestamp column (``Time (UTC)``)
-        with the load columns as data columns.
-
-    Raises:
-        FileNotFoundError: If the merged CSV does not exist.  Run
-            ``spotforecast2-entsoe download`` and ``merge`` first.
-
-    Examples:
-        ```{python}
-        import os
-        import tempfile
-
-        import pandas as pd
-        from spotforecast2_safe.configurator import ConfigEntsoe
-
-        from spotforecast2.tasks.task_entsoe import entsoe_data_loader
-
-        # Build a tiny synthetic interim CSV in a temp directory.
-        tmp = tempfile.mkdtemp()
-        csv_path = os.path.join(tmp, "energy_load.csv")
-        idx = pd.date_range(
-            "2025-01-01", periods=48, freq="h", tz="UTC", name="Time (UTC)"
-        )
-        pd.DataFrame({"Actual Load": range(48)}, index=idx).to_csv(csv_path)
-
-        # Absolute path bypasses get_data_home; loader returns the full frame.
-        config = ConfigEntsoe()
-        config.data_filename = csv_path
-        df = entsoe_data_loader(config)
-
-        print(df.shape)
-        assert df.shape == (48, 1)
-        assert df.index.name == "Time (UTC)"
-        ```
-    """
-    path = Path(config.data_filename)
-    if not path.is_absolute():
-        path = get_data_home() / path
-    if not path.exists():
-        raise FileNotFoundError(
-            f"ENTSO-E merged CSV not found at {path}.  Run "
-            "`spotforecast2-entsoe download` and `merge` first."
-        )
-    return pd.read_csv(path, index_col=0, parse_dates=True)
-
-
-def entsoe_test_data_loader(config: ConfigEntsoe) -> pd.DataFrame:
-    """Return the merged ENTSO-E CSV sliced to the forecast horizon.
-
-    The slice spans ``(end_train, end_train + predict_size * 1 h]`` so that
-    ``build_prediction_package``'s ``test_actual = ts.reindex(future_pred.index)``
-    matches the hourly forecast row-for-row.  ``end_train`` is taken from
-    ``config.end_train_default`` (treated as the *inclusive* last training
-    timestamp, the same convention the forecaster uses), and the step is
-    assumed to be 1 h after the pipeline's hourly resampling.
-
-    For the live ENTSO-E exemplar with ``end_train_default = D-2 23:00 UTC``
-    and ``predict_size = 24``, this returns the rows for
-    ``[D-1 00:00, D 00:00)`` — i.e., ``y_{-1}``.  For backtests at an arbitrary
-    ``end_train_default``, it returns the post-cutoff window the model is
-    actually predicting, rather than always "yesterday in wall-clock UTC".
-
-    Args:
-        config: A `ConfigEntsoe` with ``data_filename``, ``end_train_default``,
-            and ``predict_size`` set; the merged interim CSV must already
-            contain data covering the forecast horizon (run
-            ``spotforecast2-entsoe download`` first).
-
-    Returns:
-        DataFrame indexed by ``Time (UTC)`` with the rows the forecast will be
-        scored against.
-
-    Examples:
-        ```{python}
-        import os
-        import tempfile
-
-        import pandas as pd
-        from spotforecast2_safe.configurator import ConfigEntsoe
-
-        from spotforecast2.tasks.task_entsoe import entsoe_test_data_loader
-
-        # Synthetic interim CSV spanning the forecast window.
-        tmp = tempfile.mkdtemp()
-        csv_path = os.path.join(tmp, "energy_load.csv")
-        idx = pd.date_range(
-            "2025-12-29 00:00", periods=120, freq="h", tz="UTC", name="Time (UTC)"
-        )
-        pd.DataFrame({"Actual Load": range(120)}, index=idx).to_csv(csv_path)
-
-        config = ConfigEntsoe()
-        config.data_filename = csv_path
-        config.end_train_default = "2025-12-31 00:00+00:00"
-        config.predict_size = 24
-
-        test_df = entsoe_test_data_loader(config)
-
-        # The slice covers exactly predict_size hourly steps after end_train.
-        print(test_df.shape)
-        assert test_df.shape == (24, 1)
-        assert test_df.index[0] == pd.Timestamp("2025-12-31 01:00", tz="UTC")
-        ```
-    """
-    df = entsoe_data_loader(config)
-    end_train = pd.Timestamp(config.end_train_default)
-    if end_train.tzinfo is None:
-        end_train = end_train.tz_localize("UTC")
-    step = pd.Timedelta(hours=1)  # post-resample assumption
-    start = end_train + step  # first forecast step
-    end = start + config.predict_size * step  # exclusive upper bound
-    if df.index.tz is None:
-        start = start.tz_localize(None)
-        end = end.tz_localize(None)
-    return df.loc[(df.index >= start) & (df.index < end)]
-
-
-def entsoe_lgbm_factory(
-    config: ConfigEntsoe,
-    *,
-    weight_func: Optional[Any] = None,
-    target: Optional[str] = None,
-) -> ForecasterRecursive:
-    """LightGBM ForecasterRecursive for the ENTSO-E pipeline.
-
-    Identical to ``spotforecast2.multitask.factories.default_lgbm_forecaster_factory``;
-    kept as a named helper so the CLI's intent ("use LightGBM here") is
-    visible at the configuration site.
-
-    Args:
-        config: Any object exposing ``random_state``, ``lags_consider``, and
-            ``window_size`` (typically `ConfigEntsoe`).
-        weight_func: Per-sample weight function from the imputation step.
-        target: Ignored; accepted for factory-signature compatibility.
-
-    Examples:
-        ```{python}
-        from spotforecast2_safe.configurator import ConfigEntsoe
-        from spotforecast2_safe.forecaster.recursive import ForecasterRecursive
-
-        from spotforecast2.tasks.task_entsoe import entsoe_lgbm_factory
-
-        config = ConfigEntsoe()
-        forecaster = entsoe_lgbm_factory(config, weight_func=None, target="Actual Load")
-
-        print(type(forecaster).__name__)
-        assert isinstance(forecaster, ForecasterRecursive)
-        # The lags array is derived from lags_consider[-1] = 23.
-        assert len(forecaster.lags) == config.lags_consider[-1]
-        print("lags:", forecaster.lags)
-        ```
-    """
-    del target
-    return ForecasterRecursive(
-        estimator=LGBMRegressor(random_state=config.random_state, verbose=-1),
-        lags=config.lags_consider[-1],
-        window_features=RollingFeatures(
-            stats=["mean"], window_sizes=config.window_size
-        ),
-        weight_func=weight_func,
-    )
-
-
 def entsoe_xgb_factory(
     config: ConfigEntsoe,
     *,
@@ -235,9 +69,9 @@ def entsoe_xgb_factory(
 ) -> ForecasterRecursive:
     """XGBoost ForecasterRecursive for the ENTSO-E pipeline.
 
-    Mirrors `entsoe_lgbm_factory()` but uses an `XGBRegressor` estimator.
-    Kept as a named helper so the XGBoost variant is explicit at the
-    configuration site.
+    Mirrors `spotforecast2_safe.multitask.factories.default_lgbm_forecaster_factory`
+    but uses an `XGBRegressor` estimator.  Lives here rather than in the safe
+    package because xgboost is not a safe-package dependency.
 
     Args:
         config: Any object exposing ``random_state``, ``lags_consider``, and
@@ -251,7 +85,7 @@ def entsoe_xgb_factory(
         from spotforecast2_safe.forecaster.recursive import ForecasterRecursive
         from xgboost import XGBRegressor
 
-        from spotforecast2.tasks.task_entsoe import entsoe_xgb_factory
+        from spotforecast2.entsoe_cli import entsoe_xgb_factory
 
         config = ConfigEntsoe()
         forecaster = entsoe_xgb_factory(config, weight_func=None, target="Actual Load")
@@ -273,7 +107,7 @@ def entsoe_xgb_factory(
     )
 
 
-_FACTORY_BY_MODEL = {"lgbm": entsoe_lgbm_factory, "xgb": entsoe_xgb_factory}
+_FACTORY_BY_MODEL = {"lgbm": default_lgbm_forecaster_factory, "xgb": entsoe_xgb_factory}
 
 
 def _build_entsoe_config(model: str) -> ConfigEntsoe:
@@ -372,7 +206,7 @@ def main() -> None:
         ```{python}
         import sys
 
-        from spotforecast2.tasks.task_entsoe import main
+        from spotforecast2.entsoe_cli import main
 
         # With no subcommand, main() prints the usage summary and returns
         # without error — useful for verifying the CLI is wired correctly.
diff --git a/src/spotforecast2/tasks/__init__.py b/src/spotforecast2/tasks/__init__.py
deleted file mode 100644
index 2dbbaaf9..00000000
--- a/src/spotforecast2/tasks/__init__.py
+++ /dev/null
@@ -1,2 +0,0 @@
-# SPDX-FileCopyrightText: 2026 bartzbeielstein
-# SPDX-License-Identifier: AGPL-3.0-or-later
diff --git a/tests/test_task_entsoe_cadence.py b/tests/test_entsoe_cli_cadence.py
similarity index 70%
rename from tests/test_task_entsoe_cadence.py
rename to tests/test_entsoe_cli_cadence.py
index dfaa27b5..9629ec86 100644
--- a/tests/test_task_entsoe_cadence.py
+++ b/tests/test_entsoe_cli_cadence.py
@@ -1,14 +1,14 @@
 # SPDX-FileCopyrightText: 2026 bartzbeielstein
 # SPDX-License-Identifier: AGPL-3.0-or-later
 
-"""Tests for the 7-day retraining cadence gate in task_entsoe.train."""
+"""Tests for the 7-day retraining cadence gate in entsoe_cli.train."""
 
 import sys
 from unittest.mock import patch
 
 import pandas as pd
 
-import spotforecast2.tasks.task_entsoe as task_entsoe
+import spotforecast2.entsoe_cli as entsoe_cli
 
 
 def _make_model_file(tmp_path, project: str, age_days: float):
@@ -28,13 +28,13 @@ def _make_model_file(tmp_path, project: str, age_days: float):
 def test_train_subcommand_skips_when_recent_model_exists(tmp_path, monkeypatch):
     """A model trained 1 day ago should NOT trigger a retrain."""
     monkeypatch.setattr(
-        task_entsoe, "get_cache_home", lambda *_args, **_kwargs: tmp_path
+        entsoe_cli, "get_cache_home", lambda *_args, **_kwargs: tmp_path
     )
     _make_model_file(tmp_path, "entsoe-lgbm", age_days=1)
 
-    with patch.object(task_entsoe, "_run_entsoe_pipeline") as mock_pipeline:
+    with patch.object(entsoe_cli, "_run_entsoe_pipeline") as mock_pipeline:
         monkeypatch.setattr(sys, "argv", ["spotforecast2-entsoe", "train", "lgbm"])
-        task_entsoe.main()
+        entsoe_cli.main()
 
     mock_pipeline.assert_not_called()
 
@@ -42,15 +42,15 @@ def test_train_subcommand_skips_when_recent_model_exists(tmp_path, monkeypatch):
 def test_train_subcommand_retrains_with_force(tmp_path, monkeypatch):
     """``--force`` bypasses the cadence gate."""
     monkeypatch.setattr(
-        task_entsoe, "get_cache_home", lambda *_args, **_kwargs: tmp_path
+        entsoe_cli, "get_cache_home", lambda *_args, **_kwargs: tmp_path
     )
     _make_model_file(tmp_path, "entsoe-lgbm", age_days=1)
 
-    with patch.object(task_entsoe, "_run_entsoe_pipeline") as mock_pipeline:
+    with patch.object(entsoe_cli, "_run_entsoe_pipeline") as mock_pipeline:
         monkeypatch.setattr(
             sys, "argv", ["spotforecast2-entsoe", "train", "lgbm", "--force"]
         )
-        task_entsoe.main()
+        entsoe_cli.main()
 
     mock_pipeline.assert_called_once()
 
@@ -58,12 +58,12 @@ def test_train_subcommand_retrains_with_force(tmp_path, monkeypatch):
 def test_train_subcommand_retrains_when_no_previous_model(tmp_path, monkeypatch):
     """No saved model → retrain immediately."""
     monkeypatch.setattr(
-        task_entsoe, "get_cache_home", lambda *_args, **_kwargs: tmp_path
+        entsoe_cli, "get_cache_home", lambda *_args, **_kwargs: tmp_path
     )
 
-    with patch.object(task_entsoe, "_run_entsoe_pipeline") as mock_pipeline:
+    with patch.object(entsoe_cli, "_run_entsoe_pipeline") as mock_pipeline:
         monkeypatch.setattr(sys, "argv", ["spotforecast2-entsoe", "train", "lgbm"])
-        task_entsoe.main()
+        entsoe_cli.main()
 
     mock_pipeline.assert_called_once()
 
@@ -71,13 +71,13 @@ def test_train_subcommand_retrains_when_no_previous_model(tmp_path, monkeypatch)
 def test_train_subcommand_retrains_when_model_is_old(tmp_path, monkeypatch):
     """A model older than retrain_max_age triggers a retrain."""
     monkeypatch.setattr(
-        task_entsoe, "get_cache_home", lambda *_args, **_kwargs: tmp_path
+        entsoe_cli, "get_cache_home", lambda *_args, **_kwargs: tmp_path
     )
     _make_model_file(tmp_path, "entsoe-lgbm", age_days=10)
 
-    with patch.object(task_entsoe, "_run_entsoe_pipeline") as mock_pipeline:
+    with patch.object(entsoe_cli, "_run_entsoe_pipeline") as mock_pipeline:
         monkeypatch.setattr(sys, "argv", ["spotforecast2-entsoe", "train", "lgbm"])
-        task_entsoe.main()
+        entsoe_cli.main()
 
     mock_pipeline.assert_called_once()
 
@@ -87,4 +87,4 @@ def test_latest_saved_model_timestamp_returns_none_when_dir_missing(tmp_path):
 
     config = ConfigEntsoe()
     config.cache_home = tmp_path
-    assert task_entsoe._latest_saved_model_timestamp(config, "no-such") is None
+    assert entsoe_cli._latest_saved_model_timestamp(config, "no-such") is None
diff --git a/tests/test_tasks_entsoe_predict.py b/tests/test_entsoe_cli_predict.py
similarity index 76%
rename from tests/test_tasks_entsoe_predict.py
rename to tests/test_entsoe_cli_predict.py
index fed6268a..1e17a4d5 100644
--- a/tests/test_tasks_entsoe_predict.py
+++ b/tests/test_entsoe_cli_predict.py
@@ -6,13 +6,10 @@
 from unittest.mock import patch
 
 from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.data.entsoe_loader import entsoe_data_loader
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
 
-from spotforecast2.tasks.task_entsoe import (
-    entsoe_data_loader,
-    entsoe_lgbm_factory,
-    entsoe_xgb_factory,
-    main,
-)
+from spotforecast2.entsoe_cli import entsoe_xgb_factory, main
 
 
 def _predict_call(model: str):
@@ -21,7 +18,7 @@ def _predict_call(model: str):
     Returns ``(positional_args, keyword_args)`` of the single
     ``_run_entsoe_pipeline`` call.
     """
-    with patch("spotforecast2.tasks.task_entsoe._run_entsoe_pipeline") as mock_pipeline:
+    with patch("spotforecast2.entsoe_cli._run_entsoe_pipeline") as mock_pipeline:
         with patch("sys.argv", ["spotforecast2-entsoe", "predict", model]):
             main()
     assert mock_pipeline.call_count == 1
@@ -38,7 +35,7 @@ def test_predict_lgbm_dispatches_to_pipeline_with_predict_task():
     assert config.agg_weights == [1.0]
     assert config.bounds == [(-1e9, 1e9)]
     assert config.data_loader is entsoe_data_loader
-    assert config.forecaster_factory is entsoe_lgbm_factory
+    assert config.forecaster_factory is default_lgbm_forecaster_factory
     assert config.index_name == "Time (UTC)"
 
 
@@ -51,10 +48,10 @@ def test_predict_xgb_uses_xgb_project_and_factory():
 
 
 def test_predict_default_model_is_lgbm():
-    with patch("spotforecast2.tasks.task_entsoe._run_entsoe_pipeline") as mock_pipeline:
+    with patch("spotforecast2.entsoe_cli._run_entsoe_pipeline") as mock_pipeline:
         with patch("sys.argv", ["spotforecast2-entsoe", "predict"]):
             main()
     args = mock_pipeline.call_args.args
     kwargs = mock_pipeline.call_args.kwargs
     assert kwargs["project_name"] == "entsoe-lgbm"
-    assert args[0].forecaster_factory is entsoe_lgbm_factory
+    assert args[0].forecaster_factory is default_lgbm_forecaster_factory
diff --git a/tests/test_tasks_entsoe_train.py b/tests/test_entsoe_cli_train.py
similarity index 78%
rename from tests/test_tasks_entsoe_train.py
rename to tests/test_entsoe_cli_train.py
index a5b1640b..a394d441 100644
--- a/tests/test_tasks_entsoe_train.py
+++ b/tests/test_entsoe_cli_train.py
@@ -3,7 +3,7 @@
 
 """Tests for the ``spotforecast2-entsoe train`` CLI subcommand.
 
-Mocks ``spotforecast2.tasks.task_entsoe._run_entsoe_pipeline`` so the tests
+Mocks ``spotforecast2.entsoe_cli._run_entsoe_pipeline`` so the tests
 run offline and without needing the merged ENTSO-E CSV.  The CLI invokes
 ``_run_entsoe_pipeline(config, task="defaults", project_name=..., show=...)``
 — the config carries every pipeline parameter (targets, agg_weights, bounds,
@@ -13,13 +13,10 @@
 from unittest.mock import patch
 
 from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.data.entsoe_loader import entsoe_data_loader
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
 
-from spotforecast2.tasks.task_entsoe import (
-    entsoe_data_loader,
-    entsoe_lgbm_factory,
-    entsoe_xgb_factory,
-    main,
-)
+from spotforecast2.entsoe_cli import entsoe_xgb_factory, main
 
 
 def _train_call(model: str):
@@ -29,7 +26,7 @@ def _train_call(model: str):
     ``_run_entsoe_pipeline`` call.  ``--force`` bypasses the cadence gate so
     the dispatch test does not depend on the state of the user's model cache.
     """
-    with patch("spotforecast2.tasks.task_entsoe._run_entsoe_pipeline") as mock_pipeline:
+    with patch("spotforecast2.entsoe_cli._run_entsoe_pipeline") as mock_pipeline:
         with patch("sys.argv", ["spotforecast2-entsoe", "train", model, "--force"]):
             main()
     assert mock_pipeline.call_count == 1
@@ -46,7 +43,7 @@ def test_train_lgbm_dispatches_to_pipeline_with_defaults_task():
     assert config.agg_weights == [1.0]
     assert config.bounds == [(-1e9, 1e9)]
     assert config.data_loader is entsoe_data_loader
-    assert config.forecaster_factory is entsoe_lgbm_factory
+    assert config.forecaster_factory is default_lgbm_forecaster_factory
     assert config.index_name == "Time (UTC)"
     assert kwargs["show"] is False
 
@@ -61,17 +58,17 @@ def test_train_xgb_dispatches_to_pipeline_with_xgb_factory():
 
 def test_train_default_model_is_lgbm():
     """Omitting the positional ``model`` arg falls back to LightGBM."""
-    with patch("spotforecast2.tasks.task_entsoe._run_entsoe_pipeline") as mock_pipeline:
+    with patch("spotforecast2.entsoe_cli._run_entsoe_pipeline") as mock_pipeline:
         with patch("sys.argv", ["spotforecast2-entsoe", "train", "--force"]):
             main()
     args = mock_pipeline.call_args.args
     kwargs = mock_pipeline.call_args.kwargs
     assert kwargs["project_name"] == "entsoe-lgbm"
-    assert args[0].forecaster_factory is entsoe_lgbm_factory
+    assert args[0].forecaster_factory is default_lgbm_forecaster_factory
 
 
 def test_train_show_flag_forwards_to_pipeline():
-    with patch("spotforecast2.tasks.task_entsoe._run_entsoe_pipeline") as mock_pipeline:
+    with patch("spotforecast2.entsoe_cli._run_entsoe_pipeline") as mock_pipeline:
         with patch(
             "sys.argv",
             ["spotforecast2-entsoe", "train", "lgbm", "--show", "--force"],
diff --git a/tests/test_entsoe_pipeline.py b/tests/test_entsoe_pipeline.py
index 560fefe9..3cff8350 100644
--- a/tests/test_entsoe_pipeline.py
+++ b/tests/test_entsoe_pipeline.py
@@ -15,12 +15,10 @@
 import pandas as pd
 import pytest
 from spotforecast2_safe.configurator import ConfigEntsoe
+from spotforecast2_safe.multitask.factories import default_lgbm_forecaster_factory
 
+from spotforecast2.entsoe_cli import _run_entsoe_pipeline
 from spotforecast2.multitask.multi import MultiTask
-from spotforecast2.tasks.task_entsoe import (
-    _run_entsoe_pipeline,
-    entsoe_lgbm_factory,
-)
 
 
 def _synthetic_entsoe_df(n_days: int = 30) -> pd.DataFrame:
@@ -68,7 +66,7 @@ def entsoe_config(tmp_path: Path) -> ConfigEntsoe:
         predict_size=12,
     )
     cfg.data_loader = _make_loader(df)
-    cfg.forecaster_factory = entsoe_lgbm_factory
+    cfg.forecaster_factory = default_lgbm_forecaster_factory
     cfg.cache_home = str(tmp_path)
     return cfg
 
diff --git a/uv.lock b/uv.lock
index c91a1374..c3e6c040 100644
--- a/uv.lock
+++ b/uv.lock
@@ -3491,7 +3491,7 @@ wheels = [
 
 [[package]]
 name = "spotforecast2"
-version = "8.1.1"
+version = "9.0.0"
 source = { editable = "." }
 dependencies = [
     { name = "astral" },
@@ -3571,7 +3571,7 @@ requires-dist = [
     { name = "safety", marker = "extra == 'dev'", specifier = ">=3.0.0" },
     { name = "scikit-learn", specifier = ">=1.8.0" },
     { name = "shap", specifier = ">=0.49.1" },
-    { name = "spotforecast2-safe", specifier = ">=22.0.0,<23" },
+    { name = "spotforecast2-safe", specifier = ">=22.3.0,<23" },
     { name = "spotoptim", extras = ["torch"], specifier = ">=1.0.0,<2" },
     { name = "tqdm", specifier = ">=4.67.2" },
     { name = "ty", marker = "extra == 'dev'", specifier = ">=0.0.29" },
@@ -3590,7 +3590,7 @@ dev = [
 
 [[package]]
 name = "spotforecast2-safe"
-version = "22.2.0"
+version = "22.3.0"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "astral" },
@@ -3607,9 +3607,9 @@ dependencies = [
     { name = "statsmodels" },
     { name = "tqdm" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/5d/fd/685a4d9797d467ec646c3ffc75b8d6327ff770a7540b47c5a0300f23aac5/spotforecast2_safe-22.2.0.tar.gz", hash = "sha256:fd458aea0a6421cc8229cdbc2314b9a0863508771c286ac2537c8d3221eb1362", size = 20660329, upload-time = "2026-06-12T11:53:21.696Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/65/04/4f366dad3f78214f6d7564c883e77f67e8c734c807ac49f17c8cd1528d45/spotforecast2_safe-22.3.0.tar.gz", hash = "sha256:71ddb0215489f9042b050dda50e4eb8c0afea598ccb5c7796216073dff436e27", size = 20661753, upload-time = "2026-06-12T22:49:46.389Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/30/1b/24262db44be056e62680e4ef28e8423bd977abeac68b42c15d045f6941e3/spotforecast2_safe-22.2.0-py3-none-any.whl", hash = "sha256:1b79b7d132024103da23ccfb4e8057583e3d04906574e633c78f97e163301da9", size = 20729694, upload-time = "2026-06-12T11:53:19.078Z" },
+    { url = "https://files.pythonhosted.org/packages/1d/10/249f41e69a3b8d9803b40b721487180cea24e4d0a152fe267e1387ab0ec9/spotforecast2_safe-22.3.0-py3-none-any.whl", hash = "sha256:0791cb055e36a54deb5124bdcd8e2545ffa1ae200a89563966ac99d43891665c", size = 20731775, upload-time = "2026-06-12T22:49:43.719Z" },
 ]
 
 [[package]]