Improve analysis reliability and expand refinement workflows

.copier-answers.yml

@@ -1,6 +1,6 @@
 # WARNING: Do not edit this file manually.
 # Any changes will be overwritten by Copier.
-_commit: v0.12.0
+_commit: v0.13.1-4-g190905c
 _src_path: gh:easyscience/templates
 app_docs_url: https://easyscience.github.io/diffraction-app
 app_doi: 10.5281/zenodo.18163581

.github/workflows/docs.yml

@@ -136,11 +136,14 @@ jobs:
 
       # Build the static files for the documentation site for local inspection
       # Input: docs/ directory containing the Markdown files
-      # Output: site/ directory containing the generated HTML files
+      # Output: docs/site/ directory containing the generated HTML files
+      # Also verifies that mkdocstrings generated docs/site/objects.inv.
       - name: Build site for local check
-        run: pixi run docs-build-local
+        run: |
+          pixi run docs-build-local
+          pixi run docs-inventory-check
 
-      # Upload the static files from the site/ directory to be used for
+      # Upload the static files from the docs/site/ directory to be used for
       # local check
       - name: Upload built site as artifact
         uses: ./.github/actions/upload-artifact

.github/workflows/pypi-test.yml

@@ -38,32 +38,15 @@ jobs:
           frozen: false
 
       - name: Init pixi project
-        run: pixi init easydiffraction
+        run:
+          pixi init --platform linux-64 --platform osx-arm64 --platform win-64
+          easydiffraction
 
-      - name: Configure Pixi platforms
+      - name: Configure pixi platforms
         working-directory: easydiffraction
-        shell: bash
         run: |
-          set -euo pipefail
-
-          case "$RUNNER_OS" in
-            Linux)
-              pixi_platforms='[{ name = "linux-64-glibc-2-35", platform = "linux-64", glibc = "2.35" }]'
-              ;;
-            macOS)
-              pixi_platforms='[{ name = "osx-arm64-macos-14-0", platform = "osx-arm64", macos = "14.0" }]'
-              ;;
-            Windows)
-              pixi_platforms='["win-64"]'
-              ;;
-            *)
-              echo "Unsupported runner OS: $RUNNER_OS"
-              exit 1
-              ;;
-          esac
-
-          PIXI_PLATFORMS="$pixi_platforms" python -c 'import os, pathlib; path = pathlib.Path("pixi.toml"); text = path.read_text(); path.write_text(text.replace(text.split("platforms = ", 1)[1].split("\n", 1)[0], os.environ["PIXI_PLATFORMS"], 1))'
-          pixi workspace platform list
+          pixi workspace platform edit linux-64 --glibc 2.35 --no-install
+          pixi workspace platform edit osx-arm64 --macos 14.0 --no-install
 
       - name: Add Python 3.14
         working-directory: easydiffraction

.github/workflows/test.yml

@@ -228,28 +228,12 @@ jobs:
             echo "🔹🔸🔹🔸🔹 Python: $py_ver 🔹🔸🔹🔸🔹"
 
             echo "Initializing pixi project"
-            pixi init easydiffraction_py$py_ver
+            pixi init --platform linux-64 --platform osx-arm64 --platform win-64 easydiffraction_py$py_ver
             cd easydiffraction_py$py_ver
 
             echo "Configure Pixi platforms"
-            case "$RUNNER_OS" in
-              Linux)
-                pixi_platforms='[{ name = "linux-64-glibc-2-35", platform = "linux-64", glibc = "2.35" }]'
-                ;;
-              macOS)
-                pixi_platforms='[{ name = "osx-arm64-macos-14-0", platform = "osx-arm64", macos = "14.0" }]'
-                ;;
-              Windows)
-                pixi_platforms='["win-64"]'
-                ;;
-              *)
-                echo "Unsupported runner OS: $RUNNER_OS"
-                exit 1
-                ;;
-            esac
-
-            PIXI_PLATFORMS="$pixi_platforms" python -c 'import os, pathlib; path = pathlib.Path("pixi.toml"); text = path.read_text(); path.write_text(text.replace(text.split("platforms = ", 1)[1].split("\n", 1)[0], os.environ["PIXI_PLATFORMS"], 1))'
-            pixi workspace platform list
+            pixi workspace platform edit linux-64 --glibc 2.35 --no-install
+            pixi workspace platform edit osx-arm64 --macos 14.0 --no-install
 
             echo "Add Python $py_ver"
             pixi add "python=$py_ver"

.prettierignore

@@ -43,3 +43,4 @@ src/easydiffraction/display/plotters/assets/
 .cache
 deps/
 tmp/
+.hypothesis/

docs/dev/adrs/accepted/background-auto-estimate.md

@@ -369,7 +369,7 @@ one-line summary (§1). Empirical calibration was carried out in Phase 2:
 - The Stage-2 tolerance multiplier (`c · σ`, `c = 2`) and the width
   percentile (~75th) are first-cut constants; they were validated — not
   exhaustively swept — against the representative CWL
-  (`refine-lbco-hrpt-from-data`) and TOF (`fitting-exercise-si-lbco`)
+  (`refine-lbco-hrpt-from-data`) and TOF (`exercise-refine-si-lbco`)
   datasets plus the analytic unit cases, and produce sensible
   backgrounds there. Re-tuning stays possible if a future dataset needs
   it.
@@ -488,7 +488,7 @@ curve. **Phase 2 outcome:** the functional regression validates two
 representative datasets — CWL
 [`refine-lbco-hrpt-from-data.py`](../../../../docs/docs/tutorials/refine-lbco-hrpt-from-data.py)
 and TOF
-[`fitting-exercise-si-lbco.py`](../../../../docs/docs/tutorials/fitting-exercise-si-lbco.py)
+[`exercise-refine-si-lbco.py`](../../../../docs/docs/tutorials/exercise-refine-si-lbco.py)
 — comparing the estimated curve against the hand-placed reference to
 within a fraction of the measured signal scale; the single `arpls`
 default and the first-cut constants hold for both. Sloping and curved

docs/dev/adrs/accepted/resource-naming.md

@@ -277,7 +277,7 @@ There is no `ed-19` tutorial — id 19 is a dataset only.
 | ed-10  | PDF: Ni, NPD                                 | `pdf-ni-npd`                      |
 | ed-11  | PDF: Si, NOMAD (SNS)                         | `pdf-si-nomad`                    |
 | ed-12  | PDF: NaCl, XRD                               | `pdf-nacl-xrd`                    |
-| ed-13  | Fitting exercise: Si, LBCO                   | `fitting-exercise-si-lbco`        |
+| ed-13  | Refinement exercise: Si, LBCO                | `exercise-refine-si-lbco`         |
 | ed-14  | Refinement: Tb2Ti2O7, HEiDi                  | `refine-tbti-heidi`               |
 | ed-15  | Refinement: Taurine, SENJU                   | `refine-taurine-senju`            |
 | ed-16  | Joint: Si, Bragg+PDF                         | `joint-si-bragg-pdf`              |

docs/dev/issues/open/lowest_live-notebook-plotly-delivery-loader-vs-native-mimetype.md

@@ -33,9 +33,34 @@ gap. Two ways to fix it:
   delivery. Cons: more moving parts; depends on the notebook being
   **trusted** for a reopened (not re-run) notebook to re-render.
 
-**Current choice:** Option 2 (keeps the live-notebook extras). Revisit
-Option 1 if the loader proves fragile across notebook frontends or the
-maintenance cost outweighs the three behaviours.
+- **Colab transport — versioned frame-local assets.** Colab isolates
+  every cell output in its own iframe, so the once-per-kernel browser
+  state assumed by Option 2 cannot cross cells. In Colab only, every
+  figure output loads the version-matched Plotly bundle and
+  `ed-figures.js` from the published documentation assets in parallel
+  (the loader only needs Plotly once it renders), then calls the same
+  `renderSpec` entry point. `google.colab.output.pauseOutputUntil` holds
+  the output frame until the figure — or a visible error message — is on
+  screen. Concurrent outputs in one cell share frame-global loading
+  promises, so each asset is fetched and parsed only once. An immutable,
+  exact-release-tag jsDelivr URL is the fallback while a new
+  documentation version is deploying or if that deployment fails. Dev
+  builds intentionally use only `/dev/` to avoid mixing arbitrary local
+  code with the loader from another revision. The browser cache avoids
+  embedding or repeatedly transferring the runtime, while all
+  loader-owned theme, resize, and legend behaviour remains intact.
+  Documentation `SHARED` mode is selected first and is unaffected.
+
+  Saved Colab outputs still depend on the host restoring their dynamic
+  scripts. Until that behaviour is confirmed manually, reopen-and-view
+  without re-running is not guaranteed; re-run a cell if its restored
+  output remains at the loading placeholder. An end-to-end test of a dev
+  build likewise requires its matching loader to have reached `/dev/`.
+
+**Current choice:** Option 2 for JupyterLab, with the frame-local asset
+transport for Colab. Revisit Option 1 if the loader proves fragile
+across notebook frontends or the maintenance cost outweighs the three
+behaviours.
 
 **Depends on:** nothing.
 

docs/dev/issues/open/medium_add-mixin-tests-for-tof-size-strain-parameters.md

@@ -5,8 +5,8 @@
 **Type:** Testing
 
 The microstructural size/strain feature added four `Parameter`s to the
-TOF peak mixin (`broad_gauss_size_g`, `broad_gauss_strain_g`,
-`broad_lorentz_size_l`, `broad_lorentz_strain_l` in
+TOF peak mixin (`broad_gauss_size`, `broad_gauss_strain`,
+`broad_lorentz_size`, `broad_lorentz_strain` in
 `src/easydiffraction/datablocks/experiment/categories/peak/tof_mixins.py`),
 but they are never asserted at the mixin-test level. Their value
 hand-off is covered indirectly via

docs/dev/package-structure/full.md

@@ -305,6 +305,7 @@
 │   │   │   │   ├── 📄 factory.py
 │   │   │   │   │   └── 🏷️ class BackgroundFactory
 │   │   │   │   └── 📄 line_segment.py
+│   │   │   │       ├── 🏷️ class _EstimationArrays
 │   │   │   │       ├── 🏷️ class LineSegment
 │   │   │   │       └── 🏷️ class LineSegmentBackground
 │   │   │   ├── 📁 calculator