Open v2.4.0 planning cycle

Mark v2.3.0 as released in the completed requirements sheet and add the initial v2.4.0 planning baseline with the explicitly deferred tasks: mne-mna Python package, NeuralSet interoperability, field-map visualization, production CMNE checkpoint, and CI/release hardening follow-ups.

Author: chdinh

doc/dev-notes/v2.3.0-requirements.md

@@ -4,7 +4,7 @@ Release target: v2.3.0
 Codename: **Real-Time Ready · Cortical Studio · Python Bridge**
 Branch: staging (all work happens on staging — never push directly to main)
 
-Last updated: 13 June 2026 — All planned v2.3.0 tasks complete. Post-release-prep work (after `71a89bafd4`) added: **15+ CI pipeline fixes** (XCB/XKB deps, EGL/GLES libs, vswhere, pip cache, deploy auth, website SSG/MDX build, pipeline annotations, license-headers consolidation); **TASK 10.5** screenshot regression CI now landed (`6f74d33f98`); **TASK 11 enhancements**: session persistence for `PolhemusCoregistration` (`445fb0fc10`), constrained optical calibration solver with known tracker-to-objective distance, direct objective-center capture, sample persistence; **new disp3D features**: tracker-driven video overlay projection on head surface, vertex-displaced depth relief for video overlay, catheter probe tracking (station 3) with visualization, probe tip-only with rotating crosshair aesthetic. **TASK 4.6** field-mapping overlay + **CMNE production ONNX checkpoint** + **TASKs 1-2** deferred to v2.4.0. Tag `v2.3.0` not yet created — release pending final CI validation.)
+Last updated: 15 June 2026 — v2.3.0 released. All planned v2.3.0 tasks complete. Post-release-prep work (after `71a89bafd4`) added: **15+ CI pipeline fixes** (XCB/XKB deps, EGL/GLES libs, vswhere, pip cache, deploy auth, website SSG/MDX build, pipeline annotations, license-headers consolidation); **TASK 10.5** screenshot regression CI now landed (`6f74d33f98`); **TASK 11 enhancements**: session persistence for `PolhemusCoregistration` (`445fb0fc10`), constrained optical calibration solver with known tracker-to-objective distance, direct objective-center capture, sample persistence; **new disp3D features**: tracker-driven video overlay projection on head surface, vertex-displaced depth relief for video overlay, catheter probe tracking (station 3) with visualization, probe tip-only with rotating crosshair aesthetic. **TASK 4.6** field-mapping overlay + **CMNE production ONNX checkpoint** + **TASKs 1-2** deferred to v2.4.0. Tag `v2.3.0` created and released on 15 June 2026.)
 
 ---
 
@@ -66,7 +66,7 @@ Last updated: 13 June 2026 — All planned v2.3.0 tasks complete. Post-release-p
 | 11 | Coregistration helper app | **Done** — `mne_align` app + Polhemus FASTRAK driver + wizard + ICP + `-trans.fif` export + manual page; **session persistence** (`445fb0fc10`) + **optical calibration** enhancements landed | — | — |
 | 16 | Skigen integration & API registry | **Done** (16.7 CI workflow + validator + test all landed) | — | — |
 
-**Running totals:** all planned v2.3.0 tasks complete (36 algorithm sub-tasks + 10 GUI/infrastructure tasks); 391+ tests passing, 0 regressions. TASKs 1-2 deferred to v2.4.0. Release pending: CHANGELOG update for post-prep features, CITATION.cff date, tag creation.
+**Running totals:** all planned v2.3.0 tasks complete (36 algorithm sub-tasks + 10 GUI/infrastructure tasks); 391+ tests passing, 0 regressions. TASKs 1-2 deferred to v2.4.0. Released as `v2.3.0` on 15 June 2026; v2.4.0 planning continues in `doc/dev-notes/v2.4.0-requirements.md`.
 
 ---
 
@@ -132,14 +132,14 @@ v2.3.0 is ready to ship. All pillars are complete. TASKs 1-2 are
 deferred to v2.4.0; pillars E and F should read "Deferred to v2.4.0".
 TASK 10.5 screenshot regression CI is now landed (no longer deferred).
 
-**Release checklist (pending):**
+**Release checklist (completed):**
 
-1. Verify CI green on all platforms (staging workflows).
-2. Update `CHANGELOG.md` with post-release-prep features (video overlay, probe tracking, session persistence, optical calibration, CI fixes).
-3. Update `CITATION.cff` version to `2.3.0` and release date.
-4. Merge staging → main.
-5. Tag `v2.3.0` on main.
-6. GitHub Release with binaries triggers automatically.
+1. Verify CI green on all platforms (staging workflows). — **Done**
+2. Update `CHANGELOG.md` with post-release-prep features (video overlay, probe tracking, session persistence, optical calibration, CI fixes). — **Done**
+3. Update `CITATION.cff` version to `2.3.0` and release date. — **Done**
+4. Merge/fast-forward staging → main. — **Done**
+5. Tag `v2.3.0` on main. — **Done**
+6. GitHub Release with binaries triggers automatically. — **Done**
 
 ### TASK 16 — Skigen integration & API registry
 
@@ -170,12 +170,12 @@ inverses**, **directed connectivity**, and **MNX-as-MNE-Scan-storage**. It
 also landed the disp3D **`ElectrodeObject`** and **`SliceObject`**
 renderables — but neither is exposed in any GUI yet.
 
-As of 13 June 2026, all v2.3.0 tasks are complete. The codebase includes
+As of 15 June 2026, v2.3.0 is released. The codebase includes
 36 algorithm sub-tasks (391+ tests), all GUI pillars (MNE Inspect, MNE Scan,
 mne_analyze, mne_analyze_studio, mne_align), TASK 10.5 screenshot regression CI,
 15+ CI pipeline fixes, and new disp3D features (video overlay + catheter probe
-tracking). TASKs 1-2 are explicitly deferred to v2.4.0. The release is
-pending CHANGELOG finalization and tag creation.
+tracking). TASKs 1-2 are explicitly deferred to v2.4.0 and tracked in
+`doc/dev-notes/v2.4.0-requirements.md`.
 
 ### Pillars
 

doc/dev-notes/v2.4.0-requirements.md

@@ -0,0 +1,63 @@
+# MNE-CPP v2.4.0 — Planning Baseline
+
+Release target: v2.4.0
+Codename: **TBD**
+Branch: staging (all work happens on staging; main is reserved for releases)
+
+Last updated: 15 June 2026 — Initial planning sheet opened immediately after the v2.3.0 release. This cycle starts from the items explicitly deferred from v2.3.0 plus release-follow-up cleanup.
+
+---
+
+## Release Goals
+
+v2.4.0 should build on the v2.3.0 multimodal/real-time release by focusing on:
+
+1. **MNA Python bridge and NeuralSet interoperability** — complete the external workflow integration items that were intentionally deferred from v2.3.0.
+2. **Field-map visualization in the GUI** — expose the already-tested field-map computation stack through a scalp/BEM-aware rendering layer.
+3. **Production CMNE model packaging** — replace the in-tree fallback path with a trained, versioned ONNX checkpoint and release/update procedure.
+4. **CI/release pipeline hardening** — keep the v2.3.0 release pipeline lessons, reduce duplicate visual screenshot workflows, and make release reruns idempotent.
+
+---
+
+## Implementation Plan
+
+| Task | Description | Status | Notes / Acceptance Criteria |
+|------|-------------|--------|-----------------------------|
+| **1** | `mne-mna` Python package | Planned | Python package for reading/writing/running MNA workflows; coordinate with mne-python packaging conventions; include round-trip tests and docs. |
+| **2** | NeuralSet interoperability | Planned | Define/import/export NeuralSet event/session tables; establish conversion path to/from MNA/MNX; include schema validation and example datasets. |
+| **4.6** | Field-mapping overlay in `mne_analyze` / `mne_inspect` | Planned | `FwdFieldMap` library and tests already exist; add scalp-BEM rendering layer in `disp3D`, GUI controls, and visual tests. |
+| **3.x** | Production CMNE ONNX checkpoint | Planned | Train/version production model, document provenance, package checkpoint, and make runtime selection deterministic; keep fallback behavior for missing models. |
+| **CI.1** | Visual screenshot pipeline consolidation | Planned | Decide whether `mne_screenshot_regression` remains a separate gate with true golden-image comparison, or whether visual smoke coverage folds into doc screenshot generation (`mne_doc_shots`). |
+| **CI.2** | Release rerun hardening | In progress | Preserve idempotent release creation/tag rerun behavior; document retag procedure and expected branch/tag state. |
+| **Docs.1** | Post-v2.3 documentation cleanup | Planned | Update release-process notes with v2.3.0 lessons; ensure stable/dev website version switch remains correct. |
+
+---
+
+## Deferred / Candidate Scope
+
+### MNE-Python parity candidates
+
+Remaining non-critical gaps after v2.3.0 include:
+
+- Niche I/O formats: GDF, SNIRF, Curry, MEF, FieldTrip, Neuralynx, NSx, ANT, BCI2000, Nihon Kohden, Persyst.
+- Specialized preprocessing: OTP, PCA-OBS, fNIRS preprocessing, iEEG preprocessing, eye-tracking preprocessing.
+- Advanced forward solvers: OpenMEEG and FEM.
+- Inverse: SAM beamformer.
+- Time-frequency: Stockwell TFR.
+- Connectivity: multivariate MIC/MIM.
+- Labels: random parcellation.
+- Coregistration: MRI scaling.
+- 3D visualization: volume rendering.
+
+These are candidates for v2.4.0 only if they directly support the primary goals above; otherwise they should remain backlog items.
+
+---
+
+## Release Checklist (v2.4.0)
+
+1. Keep all work on `staging` until final release.
+2. Keep CI green across staging, screenshot/doc pipelines, CodeQL, and release dry-runs where applicable.
+3. Update `CHANGELOG.md`, `CITATION.cff`, `README.md`, and `doc/website/docusaurus.config.ts` only when v2.4.0 is ready to cut.
+4. Fast-forward `main` to `staging`.
+5. Tag `v2.4.0` on `main`.
+6. Verify GitHub Release artifacts and website update.