v0.5.0

Added

• Large-image level-of-detail (LOD) rendering: the single-image viewer now rasterises only the visible viewport at a stride matched to the current zoom, so panning and zooming very large images (4096²+) stays responsive instead of re-rendering the full array on every interaction. Profiles, statistics, histograms and hover read-outs continue to read the full-resolution data, so accuracy is unchanged.
• Memory indicator and reclamation: the menu bar shows live memory usage — both the Pyodide WebAssembly heap and the data currently held by the workspace. Clicking the indicator opens a menu gathering the memory-related actions: a Store data on disk toggle and a one-click Free memory action to reclaim memory that is no longer in use.
• Metadata editing menu: a new Edit → Metadata submenu groups the metadata actions and adds a Copy titles action (mirroring DataLab desktop's "Copy titles to clipboard").
• Delete all: a new action removes every group and object at once, behind a confirmation prompt.
• On-disk storage mode: a Store data on disk toggle (in the menu opened from the menu-bar memory indicator) spills heavy signal/image arrays to the browser's Origin Private File System (OPFS) instead of keeping them in the WebAssembly heap, so the working set is bounded by disk quota rather than the ~2 GB wasm32 memory ceiling. The HDF5 workspace remains the durable source of truth; the on-disk store is an ephemeral cache. Available only in secure contexts that support OPFS; the default stays in-memory.
• Optional worker-hosted runtime (developer-facing, opt-in via ?runtime=worker): the Pyodide runtime can run in a Dedicated Web Worker behind a typed RuntimeApi façade, moving computation off the UI thread and enabling synchronous OPFS spills (binary payloads cross the worker boundary zero-copy). The default remains the in-thread runtime, so end-user behaviour is unchanged.
• "Plot results" action: a new menu action Analysis → Plot results (available for both signals and images) aggregates scalar, statistical, or geometrical results (such as FWHM, centroid, or bounding boxes) from all selected items and plots them automatically as a new 1D signal curve, replicating DataLab desktop's behavior.
• 70 default scientific colormaps: migrated the application from 8 basic color scales to 70 exact default scientific colormaps. Smooth maps are defined using collinearity-reduced stops to keep the bundle size small (under 24 KB) and are lazily expanded into cached 256-entry RGB lookup tables at runtime.
• Directory root group naming: when using the Open from directory… feature, files residing directly in the top-level folder are now grouped in a group named after that picked folder (functional parity with DataLab desktop's path-basename fallback) instead of under a generic (root) label.
• Multi-selection group processing: aligned processing target routing on multi-selection with DataLab desktop. Processing operations on objects within groups now organize the outputs into dedicated, logically linked result groups instead of dropping them flat into the same source group.
• Multi-selection ROI extraction: aligned region-of-interest extraction on multiple selected signals or images, allowing users to apply the current object's ROI to all selected objects and grouping the outputs cleanly.
• Blob detection grouping: grouped blob detection under the Analysis submenu to mirror the DataLab desktop menu organization.

Changed

• Menu and toolbar icons are now embedded directly in the bundle instead of being fetched as separate files, so they appear instantly on first paint without a brief flicker and without extra network round-trips.
• Optimized bundle splitting and lazy-loading: reduced the initial app bundle footprint from 2.05 MB to ~414 kB gzipped. Heavier vendor dependencies (Plotly, CodeMirror, and React) are now split into dedicated cacheable chunks, and code-editor panels (Macro and Notebook panels) are lazy-loaded after first boot.
• Fast binary signal serialization: large signal data arrays are now serialized using raw binary buffers via zero-copy views (Float64Array) instead of JSON stringification across the JS/Pyodide boundary, dramatically reducing serialization processing overhead.
• Offscreen rendering and OPFS paging latency: parallelized OPFS page-in disk operations via concurrent promises and optimized offscreen canvas paint loops (avoiding redundant pixel-level color allocations), ensuring responsive performance during heavy viewport updates.

Fixed

• Changing a displayed object's properties (title, axis labels/units…) from a macro, notebook, remote-control call or the AI assistant now refreshes the central plot immediately, instead of updating only the object tree and leaving stale axis titles/units on the graph.
• Fixed spurious get_image_data failed console errors when a script or notebook creates a signal and then an image in quick succession (e.g. the Signal & image processing notebook template): a panel refresh could briefly leave the image viewer pointed at a signal object. The processing results were unaffected; only the stray errors are gone.
• Running a notebook that switches between the signal and image panels mid-run (e.g. Run all on a template that creates a signal and then an image) no longer leaves the central viewer pointed at the wrong object kind or raises stray errors.
• Editing an image's properties (title, labels, units…) no longer triggers a console error; image metadata edits now refresh the image view correctly instead of attempting to read it as a signal.
• In Store data on disk mode, the object tree now reports each object's real size (number of points, width × height) instead of showing "1 pt" for every object whose data has been spilled to disk.
• Region-of-interest extraction (signal and image) and the image Erase area action now produce fully resolved result titles, instead of leaving unresolved placeholders such as extract_roi({0}).
• Deleting an image group no longer fails with an "Unknown group" error.
• Auto-generated parameter dialogs now show the correct heading, taken from the dataset's own title.
• Updated the Microsoft Edge "slow load" hint: it now points at Edge's secure-mode site settings and explains how to allow DataLab-Web to run at full speed.
• Dynamic active-state grayout: implemented full active-state tracking (active_state parameter) across all generated guidata forms, dynamically graying out inactive or conditionally disabled parameter fields in real time.
• Computed read-only field rendering: fixed styling and visibility of computed or read-only guidata parameters (such as the dynamic "Operation" field in the Arithmetic dialog), displaying them as clear, immutable fields in the side panel.
• Complete French translation parity: added dynamic menu and analysis feature keys from Sigima into the locales extraction registry and fully translated them to French, bringing translation parity with DataLab Qt.
• Cursor drag mode on image stats: activating the image statistics tool now automatically forces the Plotly cursor dragmode to rectangular selection (drawrect), removing the need to manually switch modes to define the ROI.
• Radial profile processing flow: aligned the image radial profile dialog, profile center parameters, interactive behavior and auto-titles with the desktop application.
• Simultaneous 1-to-0 analyses: 1-to-0 analyses (such as FWHM, centroid, etc.) now run on all selected signals concurrently instead of just the last selected one, drawing merged geometry overlays on the plot and formatting result tables to match DataLab desktop's columns (e.g., contracting segment analyses to single dX columns).
• Multi-selection action enablement: actions like Remove all ROIs and Clear results are now unlocked when any of the selected objects meet the criteria, operating across the entire selection.
• Multi-image grid loop guard: resolved React key collisions and infinite render loops occurring on the multi-image grid panel when list groups had duplicate object IDs.
• Active tab color: aligned the background color of selected Signals/Images tabs with the panel background color for a fully integrated visual design.

Full Changelog: v0.4.0...v0.5.0

v0.4.0

Added

• Internationalisation framework: the UI now renders in the user's regional language (auto-detected, with a language selector in the menu bar and a ?lang= URL override). English is the source language and French is the first translated locale.
• Spreadsheet array editor: signal and image raw data (signal X/Y values, image pixel matrices) can now be edited directly from the properties panel through a spreadsheet-style array editor dialog, with an enriched array preview.

Changed

• Inactive and computed parameter fields are now displayed read-only in the properties panel, matching DataLab Qt's behaviour.

Fixed

• Non-uniform images now render with their exact pixel coordinates (correct extent and hover Z values) instead of being collapsed onto a uniform grid.
• Picture-in-Picture floating windows and their pagehide listener are now properly cleaned up when the side panel is closed.
• Pending annotation writeback is cancelled when a signal plot is unmounted, avoiding writes to stale objects.
• Legacy AI Assistant API keys are migrated to encrypted storage without creating duplicate entries.

Security

• guidata field labels and descriptions are now sanitised before being injected as HTML in auto-generated parameter forms.
• The HDF5 browser tree now guards against unbounded recursion depth when exploring deeply nested files.

Full Changelog: v0.3.0...v0.4.0

v0.3.0

Full Changelog: v0.2.0...v0.3.0

v0.2.0

Full Changelog: v0.1.0...v0.2.0

v0.1.0

Full Changelog: https://github.com/DataLab-Platform/web/commits/v0.1.0