chapkit 2.1.0__tar.gz → 2.1.1__tar.gz

{chapkit-2.1.0 → chapkit-2.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: chapkit
-Version: 2.1.0
+Version: 2.1.1
 Summary: ML and data service modules built on servicekit - config, artifacts, and ML workflows
 Keywords: ml,machine-learning,data-science,artifacts,config,servicekit
 Author: Morten Hansen

{chapkit-2.1.0 → chapkit-2.1.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "chapkit"
-version = "2.1.0"
+version = "2.1.1"
 description = "ML and data service modules built on servicekit - config, artifacts, and ML workflows"
 readme = "README.md"
 requires-python = ">=3.13"

{chapkit-2.1.0 → chapkit-2.1.1}/pyproject.toml.orig

@@ -1,6 +1,6 @@
 [project]
 name = "chapkit"
-version = "2.1.0"
+version = "2.1.1"
 description = "ML and data service modules built on servicekit - config, artifacts, and ML workflows"
 readme = "README.md"
 authors = [{ name = "Morten Hansen", email = "morten@dhis2.org" }]

{chapkit-2.1.0 → chapkit-2.1.1}/src/chapkit/api/apps/console/manifest.json

@@ -1,6 +1,6 @@
 {
   "name": "Chapkit Console",
-  "version": "2.1.0",
+  "version": "2.1.1",
   "prefix": "/",
   "description": "Default web console for exploring and operating a chapkit service",
   "author": "Chapkit Team",

{chapkit-2.1.0 → chapkit-2.1.1}/src/chapkit/cli/test/command.py

@@ -415,6 +415,7 @@ def test_command(
                     historic, future = generator.generate_prediction_data(
                         num_locations=5,
                         num_periods=future_periods,
+                        historic_periods=max(1, num_predict_rows // 5),
                         required_covariates=runner.required_covariates,
                         additional_covariates=pred_acc,
                         extra_covariates=extra_covariates,

{chapkit-2.1.0 → chapkit-2.1.1}/src/chapkit/data/generator.py

@@ -155,12 +155,15 @@ class TestDataGenerator:
         additional_covariates: list[str] | None = None,
         extra_covariates: int = 0,
         period_type: Literal["monthly", "weekly"] = "monthly",
+        historic_periods: int | None = None,
     ) -> tuple[dict[str, Any], dict[str, Any]]:
-        """Generate historic and future DataFrames for prediction."""
+        """Generate historic and future DataFrames for prediction; historic_periods defaults to num_periods."""
+        if historic_periods is None:
+            historic_periods = num_periods
         # Historic data: actual observations from earlier time periods
         historic = self.generate_training_data(
             num_locations=num_locations,
-            num_periods=num_periods,
+            num_periods=historic_periods,
             num_features=num_features,
             required_covariates=required_covariates,
             additional_covariates=additional_covariates,
@@ -170,7 +173,7 @@ class TestDataGenerator:
         )
 
         # Future data: continue immediately after the historic span (same origin year,
-        # offset by num_periods) so the future horizon is contiguous with history
+        # offset by historic_periods) so the future horizon is contiguous with history
         # instead of jumping to a fixed year and leaving a calendar gap. Then null out
         # disease_cases for prediction.
         future = self.generate_training_data(
@@ -181,7 +184,7 @@ class TestDataGenerator:
             additional_covariates=additional_covariates,
             extra_covariates=extra_covariates,
             start_year=2020,
-            start_period=num_periods,
+            start_period=historic_periods,
             period_type=period_type,
         )
 

{chapkit-2.1.0 → chapkit-2.1.1}/src/chapkit/ml/router.py

@@ -117,10 +117,12 @@ class MLRouter(Router):
 
             if kind == "predict":
                 # The future frame's period count is the horizon the service will resolve
-                # at $predict, so keep it inside the bounds the service declares.
+                # at $predict, so keep it inside the bounds the service declares. The
+                # historic frame keeps the full requested length as model context.
                 historic, future = generator.generate_prediction_data(
                     num_locations=num_locations,
                     num_periods=_clamp_horizon(num_periods, sample_metadata),
+                    historic_periods=max(1, num_periods),
                     num_features=num_features,
                     required_covariates=required_covariates,
                     period_type=resolved_period,