Site navigation

Try in browser (opens in a new tab)
Download VarvePublic Beta
Release historyv0.2.1Beta

Varve changelog

Every release of Varve, documented. Features, fixes, and improvements drawn from the project's release notes.

Download Varve GitHub releases

  1. #0.5.0

    Release preview

    Prepared source-build notes; installers remain v0.2.1.

    Added and improved

    • Six task workspaces and flexible panels — Design, Print, Draw, Photo, Motion, and Email share one document model. Docked, floating, and split panel layouts can be adjusted and restored; Logo tools stay in Design and Code export remains a shared panel.
    • Illustration and tablet controls — Separate flats and clipped shading layers, brush opacity and flow, pressure settings, and one-finger draw or navigation controls extend the painting workflow. Check pressure and tilt support on the device you use.
    • Comic lettering — Build speech, thought, and shout balloons, captions, panel layouts, and editable lettering with localization support. Pointed balloon tails now join the bubble outline without a line across the pointer; existing balloons are corrected when their documents migrate.
    • Photo workflows — Source-preserving retouch, Frequency Separation, Liquify, depth-aware masks, bracket merging, and verified Ultra HDR sharing join image tuning and colorization. RAW development is limited to the documented camera subset; model-based features may require optional downloads.
    • Vector and type tools — Shape Builder, compound-path editing, more precise tracing, path text, variable-font controls, and shaped-glyph outlines improve illustration and lettering workflows.
    • Reusable pattern authoring — Create definitions from copied vector art, raster tiles, or recipes; tune repeat and per-fill placement, inspect raster seams, and edit copied vector motifs in a bounded source editor. Imported raster tiles now remain rendered alongside other document images. SVG and PDF support defined subsets; unsupported PDF layouts are warned and omitted.
    • Presentation decks — Order editable Design frames, preview a deck, and export slides as a raster screen PDF or numbered PNG archive. PDF output is not editable/searchable and has no tagged accessibility structure; layout formatting inheritance remains limited.
    • Manual Token Sync — Import and re-import token sources and review mapping diagnostics. Automatic watching, Git integration, and write-back to source files are not included.
    • Local Wasm plugins — Review and run bounded local packages for selected- layer analysis and batch renaming. There is no marketplace, arbitrary JavaScript/HTML, or plugin-owned network and filesystem access.
    • Mockups and exports — Folded-mesh mockups join planar, perspective, and cylindrical surfaces. SVG and PDF export preserve more supported artwork; unsupported content can still use raster fallbacks or documented warnings.
    • Experimental rendering options — Canvas2D remains the default. WebGL2 and WebGPU are opt-in experiments with capability checks and Canvas2D fallback; results depend on the browser, driver, and hardware.
    • Document compatibility — The app now writes document schema 2.33; published v0.2.1 used schema 2.21. Older documents migrate on open, including the pointed-balloon outline correction, but newer files are not backward-compatible. Back up important .varve files before opening them.
    • Save, history, and recovery — Saved, Saving, Modified, and Save failed states are explicit, and failed writes retain recovery paths. Confirm a save by its successful destination write before closing important work. Buffered text edits and formatting now replay through undo/redo with exact grapheme boundaries, including emoji and combining characters.
    • Responsive and accessible controls — Workspace panels, menus, touch targets, focus handling, home-screen reflow, group dragging, and enlarged-text layouts have been refined across compact, tablet, and desktop widths.
    • Clearer selection and import feedback — Spacing labels avoid selected artwork, and missing-font recovery opens after the import summary is closed.
    • Bounded browser demo — /try/ remains a sample-document experience with browser-local storage and a limited workspace set; desktop-only print, inference, and other capabilities are not implied by the demo.

  2. #0.2.1

    LatestBeta
    View on GitHub

    Added

    • Image Enhance — batch enhancement processes every image in a multi-selection in one run, with combined tile progress, per-image announcements ("Enhancing image 2 of 3…"), immediate cancellation, and a failure summary instead of stopping silently mid-queue. A validated Real-ESRGAN Anime x4 (6B) ONNX model (checksum-pinned download) now backs Illustration mode, restoring preview/apply parity for that mode. Deblur strength is user-controllable, the dialog reports real per-stage progress while restoration runs, and output stays non-destructive with a quality policy (faithful/balanced) choice.
    • Canonical export resolution model — export scale decisions are now computed by one shared resolution engine in @varve/scene (export/resolution, documented in docs/architecture/export-resolution.md), and batch export exposes explicit resolution controls.
    • Text on Path becomes reachable — the engine could always place type along a path, but nothing in the UI ever set the required fields. Object ▸ Text on Path now attaches a selected text layer to a selected shape, and a Text on Path inspector section exposes the start offset and which side of the curve the glyphs sit on.
    • Linux download guidance — the website download page adds a distro-family picker table, grouped install commands, expanded Arch/CachyOS troubleshooting, and per-format system requirements, and now recommends the AppImage as the primary Linux format.
    • Verified workflow recordings — a deterministic capture pipeline (scripts/capture/) records real interactions against seeded demo documents, verifies them frame-by-frame, and publishes the resulting workflow videos to the website product page.
    • Selection tools — a pixel-lasso area-selection tool, a quick-mask paint tool, and image-derived selections join rectangle/ellipse marquee selection. Selections can be transformed, refined (feather, contract/expand, smooth), and saved for reuse across a session, with a dedicated panel for managing saved selections and reviewing where the current selection came from.
    • Layer States — layers gain a Solo View toggle and per-layer visibility state that survives independently of the old show/hide toggle, with an effects badge and layer context menu to manage them.
    • Perspective tool — an interactive four-corner perspective transform with a live overlay, correct Canvas2D and SVG-export rendering, and its own Inspector section.
    • Smart Filters and adjustment layers — a non-destructive filter stack can be applied per-object, with menu and command-palette entries and an object filter stack that respects layer-level bypass.
    • Shadow/Highlight adjustment, Levels/Curves editors backed by real histogram data, and general adjustment-layer normalization fixes.
    • Crop and transform — named crop aspect ratios, crop guides, straighten, repeat-transform, and a rebuilt Image Resize dialog.
    • Find Similar — natural-language and image-based asset search in the Intelligence panel, backed by a local DINOv2 embedding pipeline.
    • Warp tool is now reachable from the Design toolbar and its W keyboard shortcut; it previously existed but had no way to activate it.
    • Accessibility — a full audit pass across the editor (focus order, ARIA roles/labels, keyboard reachability), plus a specific fix reversing Tab order so it flows from the layers panel into the canvas instead of skipping it.

    Changed

    • Git LFS is no longer used anywhere in the repository (bandwidth budget exhausted): all media is committed directly and CI no longer fetches LFS objects.
    • The updater feed generator (scripts/release/generate-updater-feed.mjs) now supports Linux AppImage and Windows NSIS targets for both x86_64 and ARM64, and the macOS .app.tar.gz target when a signed bundle exists, verifying signatures cryptographically before publication. The live v0.2.0 feed (apps/website/public/updates/stable.json) was re-signed and republished for its Linux x86_64/ARM64 AppImage targets only — Windows and macOS artifacts were not re-signed as part of that republish, so the feed does not yet carry windows-* or darwin-aarch64 entries.
    • Release version stamping also updates AUR packaging metadata and AppStream metainfo.
    • ONNX Runtime failures report the platform-specific remedy instead of a generic message, and the runtime library is staged before every desktop build path.

    Fixed

    • Variable fonts apply their axes — fvar parsing matches how OpenType actually stores the table, the selected wght value reaches the Canvas2D painter on every text path, variation settings survive onto the painted shape, and the variable-axis panel is reachable without scrolling the inspector sideways.
    • Pixel-art scaling at non-power-of-two sizes — EPX/HQx/xBR no longer splice a nearest-neighbor remainder into the middle of the algorithm at scales like 3x; the 2x passes stay intact.
    • Timeline drag preserves keyframe progress during the drag gesture.
    • Prototype, codegen, variants, and timeline workflows were repaired where editor action registration had broken them, and text-on-path actions see live selection state.
    • Image Enhance reliability — tiled inference routes to the right capability, tiled recomposition memory is bounded, edge tiles produce correct output dimensions, stale previews clear when the operation changes, stale-result errors surface to the UI instead of returning silently, and the Auto analysis no longer produces NaN JPEG-blockiness scores. Compression-artifact removal remains unavailable by design: the dialog says so rather than degrading another model.
    • Visual constraint editor — the Constraints inspector section rendered a passive preview even though the interactive pin-control component already existed; it's now wired up and usable. Its horizontal/vertical/center-stretch pin targets previously overlapped in the same hit area so only the last one was clickable — they're now separate targets.
    • Drawing a frame around an existing image no longer requires the image to be fully inside the frame before it's adopted — the original center-hit-or-≥50%-overlap rule is restored.
    • Multi-page documents — objects created on any page after the first had the page's canvas placement applied twice, landing (and hit-testing) in the wrong position; fitting or viewing a non-first page also ignored its actual placement and always fit at world origin. Both are fixed.
    • Mockup creation selected a placeholder ID instead of the frame node it had actually just created, so the Inspector's mockup section never appeared after applying a mockup to a source image.
    • Imported LUTs (.cube) were applied at zero opacity, so an imported LUT adjustment never visibly did anything until manually adjusted.
    • Icon library — inserting an icon and immediately switching to the Downloaded filter could miss the icon just inserted, due to a cache-refresh race; the panel now reflects an insert immediately.
    • Typing or pasting large amounts of text could crash the editor with a React "maximum update depth exceeded" error in the text tool; rapid keystrokes are now coalesced into bounded updates, while blur, Escape, and IME completion still flush immediately.
    • Cross-artboard drag reparenting and its undo/redo now land at the precise expected position.
    • Edge auto-pan while dragging near a canvas boundary moved content in the opposite of the intended direction; the pan direction is corrected.
    • Help ▸ Getting Started did nothing when clicked; it now opens onboarding.
    • Responsive floating toolbar — tools that don't fit the current window width (including Table) now live in an accessible "More tools" overflow menu instead of being unreachable.
    • Crop tool history — committing a crop without changing anything no longer pushes an undo step, so Undo after an untouched crop correctly reverts the prior action instead of the no-op crop.
    • Depth Blur now renders through the native/WASM engine path — the effect existed only in the Canvas2D/JS layer, so the native scene representation didn't recognize it and every use silently took the slower fallback path.
    • Background removal (INT8/fast mode) — the small-download and low-memory variant of the U²-Net Light model produced a near-blank mask; a runtime safety check already kept it from ever being served, so this never reached a shipped build, but the fast/small-download option is now actually usable instead of silently falling back to the full-precision model.

    Security

    • The basic-auth-URL secret scanner no longer false-positives on JSON-LD structured data.

  3. #0.2.0

    Beta
    View on GitHub

    Added

    • Motion/prototyping P1-P3 improvements — timeline virtualization (only visible tracks rendered), Lottie fill/stroke color keyframe export, motion path drag-to-edit on canvas, prototype startAnimation/stopAnimation playback wiring, prototype interaction section UI for Play/Stop animation actions, and prototype click-through E2E test. 413 tests pass across 31 files.
    • Figma REST JSON import — official Figma file JSON (REST API or plugin export) is now a first-class import source. A bounded source normalizer (figma/source.ts) enforces 64 MB / 100k node / 256 depth limits, then a semantic converter (figma/converter.ts) maps pages, frames, groups, shapes, text, Auto Layout, components/instances, variables, styles, and prototype interactions into native Varve document fragments with fresh IDs and deduplicated image assets. Opaque .fig binaries are rejected with actionable guidance. Fidelity is honest: boolean operations, remote image refs without embedded data, and unsupported effects are reported rather than silently dropped. See docs/architecture/figma-import-system.md for the full conversion matrix and architecture.
    • Image Enhance — Deblur and Auto/Recommended — the Enhance workflow now ships a validated Deblur operation backed by a reproducible conversion of NAFNet-GoPro-width64 (MIT, ~138 MB fp16 ONNX, downloaded on demand through the model manifest with a pinned checksum; conversion parity is bit-exact with the trusted PyTorch reference — 98.9 dB on the official GoPro test subset). Deblur runs through the same shared native→worker provider chain as Denoise, with adaptive tiling that keeps tiles single-shot up to 1280 px because NAFNet's global receptive field makes small tiles visibly seamed. The dialog opens in Auto/Recommended mode: a cheap classical analysis (noise, blur, JPEG blockiness, resolution) proposes a restoration in human terms with a confidence number, and Restore + Upscale composes only the stages it needs (denoise, deblur, or deblur+upscale, always restoration before super-resolution).
    • Denoise fix: graph-safe padding — the SCUNet ONNX conversion actually requires padded dimensions divisible by 64 (its baked attention reshape crashes otherwise), so 1080p and other non-64- multiple images failed denoise. Padding, the native spec, and the manifest contract were corrected (previously claimed 8).
    • Restore benchmark tooling — scripts/bench/restore-reference/ provides a deterministic degradation corpus (JPEG, Gaussian, motion blur recipes with fixed seeds), TS-exact ONNX reference runners, and contact-sheet rendering; measured results are in docs/quality/image-enhancement-benchmark.md. Compression-artifact removal remains unavailable by design: no model passed the design-content corpus (SCUNet destroys 1px line patterns; the only JPEG-trained NAFNet checkpoint was rejected on provenance).
    • Natural-language asset search — the Asset Browser search field now combines filename, OCR, tags, and metadata with an optional local visual lane. Describe what you remember ("orange sunset over mountains") and matching local assets rank by visual content even when the file is named IMG_4281.jpg. Images are indexed in the background with a bounded, cancellable queue (deduplicated by content hash, so renames and copies never re-embed), search results keep match reasons, and exact filename queries keep their ordering guarantee. The text tower and tokenizer are parity-verified against the reference implementation; everything runs locally with no uploads. Visual search is opt-in: the SigLIP image and text models plus tokenizer download explicitly, verify SHA-256, and filename/OCR/metadata search keeps working without them. See docs/architecture/asset-search-system.md and ADR-0221.
    • Object Selection — select an image and use the Select Object tool (or the Adjustments tab) to prompt a local SAM2-Hiera-Tiny model with positive/negative points and drag boxes. The preview is transient until you apply it: candidate masks can be cycled before applying, and Apply creates one undoable non-destructive raster mask that survives save, reload, and model removal. The ~155 MB model is an explicit, checksum- pinned download (Apache-2.0) loaded lazily; embeddings live in a bounded session cache and are never written into the document. Promptable segmentation is not semantic subject detection and is not a perfect alpha matte — brush and trimap refinement remain the edge-quality tools.
    • Depth-aware effects — a reusable, model-independent DepthMap resource powers non-destructive Depth Blur: pick a focus point, adjust focus range and blur strength, preview the depth field, or convert a depth range into a layer mask. Depth maps are generated on demand by a ~27 MB local model (Depth Anything V2 Small, Apache-2.0, SHA-256 pinned), cached per source revision, persisted in the document (16-bit scalar field), and rendered without the model, so saved documents reopen with identical results and no inference. The blur compositor is occlusion-aware (sharp subjects do not smear into blurred backgrounds) and premultiplied-alpha correct. Relative depth only; no metric calibration is claimed.
    • Experimental asset similarity — the Intelligence panel now separates image-to-image Similar search from Near duplicates. The image lane uses a local DINOv2-small encoder (Apache-2.0, SHA-256 pinned, reference-vector parity verified against an independent runtime; selected over the SigLIP image encoder from a Varve-corpus evaluation — see docs/audits/semantic-asset-similarity-evaluation-2026-08-13.md). Near-duplicate ranking keeps exact identity and perceptual fingerprints separate. Computed embeddings are cached locally by content hash, so unchanged images never re-run inference. The current workflow is document-local, capped at 30 image candidates, and does not provide automatic deletion. See docs/architecture/semantic-asset-similarity.md.
    • Image palette extraction — select one image and open Appearance → Palette to generate a deterministic local palette in perceptual Oklab, review generated harmonies and WCAG 2.1 contrast pairs, copy HEX values, and save extracted colours as document swatches or colour variables. Analysis is bounded, cancellable, worker-backed when available, and does not upload image pixels or add derived analysis data to the document schema.
    • Email workspace (desktop) — a new workspace mode (Ctrl+Shift+6) for visual email authoring with a dedicated IR, HTML and plain-text compilers, embedded-asset packaging, URL preflight diagnostics, and multi-provider output. Template types, preview, and export are available today; rendering fidelity depends on the recipient's email client.
    • Browser demo — a bounded public demo at /try/ runs a curated sample poster document in WASM with honest capability messaging, stale-asset recovery, and a desktop CTA. See docs/architecture/browser-demo.md.
    • Auto-layout improvements — drag-to-reorder children within a flex frame, "Add Auto Layout" command to wrap a selection, per-axis child sizing (width/height, fit/hug/absolute), grid hug sizing, and double-click a resize edge to reset that axis to Hug.
    • OS file associations — .varve and .strata files register with the operating system's "Open With" on Linux, macOS, and Windows so double-clicking a document opens it in Varve.
    • Onboarding — canvas empty-state shortcuts, micro-hints for new tools, learning preferences, and a What's New dialog that surfaces recent changes.
    • Workspace mode redesign — the mode picker now uses per-mode accent colours, elevated depth, and improved hover states for faster switching.
    • Contact channels — canonical contact and security surfaces in both the application and the website (support@varve.studio, security@varve.studio, GitHub Private Vulnerability Reporting).
    • Paint tools — raster brush system with a Brush Browser (portable brush packages, deterministic previews) and Brush Editor (size, flow, hardness, spacing, scatter, texture, grain). Wet-media lifecycle (wet-edge blending, sample-all-layers clone), real smudge transport with vector-pressure support, symmetry guides, clone source markers, alpha lock, and mask painting on raster layers. The paint inspector wires presets and per-stroke settings to the worker pipeline. See docs/architecture/paint-system.md.
    • Gradient map adjustments — a non-destructive gradient-map adjustment layer remaps tonal values through a user-editable colour gradient with a built-in preset browser. Affects vector and raster content; export preflight flattens to raster for SVG/PDF targets. See packages/scene/src/gradientPresets.ts for the preset model.
    • Selection improvements — object marquee with configurable shape modes (rectangle, ellipse, lasso), area-aware selection commands (select all in area, deselect, toggle), hierarchy navigation (select parent/child), select-similar expansion, and indexed geometry for fast broad-phase overlap. Raster masks can now be bridged from area selections.

    Changed

    • Windows installer shrunk ~75% — the NSIS installer now bootstraps WebView2 via the Microsoft-provided downloadBootstrapper instead of bundling it, and the bundled ORT WASM runtime was trimmed.
    • Licensing — eight engine crates (varve-core, varve-colour, varve-trace, varve-layout, varve-media, varve-effects, varve-upscale, varve-bgremove) are now published under **MIT OR Apache-2.0** for ecosystem reuse and grant eligibility.
    • Brand cleanup — remaining Strata references in export identifiers, wordmark SVGs, product screenshots, and website copy have been corrected to Varve.
    • Render resource management — adaptive image fidelity policy and pressure-signal integration improve frame pacing under memory pressure.
    • Marketing website — comparison page, feature screenshots, SEO metadata, download-page troubleshooting, and download funnel with platform recommendation.

    Fixed

    • Accessibility — dialog touch targets, focus-visible outlines, keyboard navigation in RecoveryDialog, aria-disabled on inactive buttons, and error-boundary auto-focus.
    • Canvas performance — getBoundingClientRect is now cached in the input-pipeline hot path instead of called on every pointer-move.
    • Layout engine — flex bugs repaired (recursive hug sizing, per-axis child sizing, frame own-size reflow, grid hug sizing).
    • Motion playback — hardened sampler and prototype runtime against edge cases that could stall or crash.
    • Popover and Select — portaled custom Select no longer leaks outside-click closures; submenus position correctly.
    • Multi-selection — Ctrl+drag preserves the existing selection when the target is already selected.
    • Panel styling — previously unstyled surfaces (ErrorBoundary, AdjustmentEditor, Design Audit panels, VariablePanel headers) now receive consistent design-token styling.
    • Branding — stale .strata references replaced with .varve in user-facing code and current-state docs; export identifiers no longer emit Strata names.
  4. Show 3 older releasesHide 3 older releases

    1. #0.1.2

      Beta
      View on GitHub

      Fixed

      • Settings dropdowns now render above native dialogs and Export, Performance, Appearance, and theme changes apply immediately.
      • Content-Aware Fill preview sizing and Fit, 1:1, Center, zoom, keyboard, and crop interactions now behave consistently.
      • Palette extraction accepts a user-selected count from 3 through 32 colors.
      • Browser update synchronization, Linux browser chrome detection, menubar navigation, and React 19 tooltip refs are hardened.
      • GitHub Actions now have stronger workflow validation, concurrency controls, cache/log diagnostics, and failure reports.
      • GitHub Pages builds use the correct /varve/ project-site base path.

    2. #0.1.1

      Beta
      View on GitHub

      Fixed

      • AppImage white screen on modern Mesa/EGL hosts: the bundled WebKit/GTK closure (built on the ubuntu-22.04 baseline) fails EGL display creation on newer Mesa stacks. scripts/release/prune-appimage-bundled-libs.mjs now strips usr/lib from the AppImage payload during the release build — the binary resolves everything from the host, verified on a CachyOS host where the released AppImage aborted and the pruned one rendered correctly. The AppImage now requires system WebKitGTK (libwebkit2gtk-4.1) like the .deb; the download page documents this instead of promising "runs on any Linux". (The earlier in-app WEBKIT_DISABLE_DMABUF_RENDERER=1 workaround was kept but is not sufficient on its own: the EGL failure precedes renderer selection.)
      • The release launch smoke can no longer pass on a blank window: it fails on the EGL/abort signature in the app output and requires a live WebKitWebProcess.
      • The release draft job could never create a draft (v0.1.1 rehearsal): the files glob used a negated !dist/release/RELEASE_NOTES.md pattern, which softprops/action-gh-release's npm-glob matching treats as matching nothing, so fail_on_unmatched_files: true aborted the job. RELEASE_NOTES.md is now staged outside the globbed directory and the files list is a single positive pattern; scripts/validate-workflows.mjs rejects any future reintroduction of the negated pattern or a body_path inside the glob.
      • Release hardening inherited from the 0.1.0 rehearsal: frontend built before desktop compilation, Git LFS fetched on every checkout (bundled models are LFS-tracked), LFS-pointer guard fixed for Windows paths, per-platform SBOMs generated and validated, SHA256SUMS.txt generated last over the complete upload set, draft assets re-downloaded and re-hashed before publication, native runner smokes for Windows and macOS, and a container install-test for the Linux packages.

      Added

      • Code-signing pipeline (certificate-ready) — the release system now enforces a fail-closed signing policy: a signing-preflight job validates Apple/Azure credentials before any build starts, Windows installers are signed through Azure Artifact Signing via Tauri's signCommand, macOS builds are Developer ID signed, notarized and stapled via the App Store Connect API, and every artifact is verified on its actual bytes (verify-windows-signature.ps1, verify-macos-signature.sh) before checksums, GitHub artifact attestation, and the draft release. Signedness in release metadata derives only from those verification reports — a stable release never silently ships unsigned. See docs/release/signing-decision-record.md for the strategy, docs/release/code-signing-setup.md for the human acquisition checklist. No certificates are owned yet; until they are, releases are unsigned and honestly labelled.
      • Image Trace (native raster-to-vector) — trace a selected image into editable vector artwork with presets (crisp logo, pixel-art sprite, centerline sketch and more), live preview, and a result estimate. Desktop builds run the native Rust engine on a background thread with progress and cancellation; web builds fall back to bounded TypeScript providers with honest capability gating (centerline is native-only). Modes: black-and-white outline (threshold, despeckle, holes), grayscale, limited color (perceptual Oklab palette), pixel art (hard pixel boundaries, nearest-neighbor scaling), and centerline (stroked skeletons). Traces insert as one undoable group beside the source; Edit Trace (context menus) re-opens with the stored settings and replaces the result in place. Entry points: Object menu, canvas/layers context menus, command palette (Ctrl+Alt+Shift+T), Inspector, and QuickBar. See docs/architecture/image-trace-system.md.
      • New Design experience — the New File dialog is now "New design": an editable document name (untitled defaults with collision-free numbering), three starting points (Empty document / Start with a frame / Template), searchable frame presets with favorites and recents, custom frame sizes with unit conversion and aspect lock, and collapsed Advanced settings (print intent reveals CMYK, DPI, and bleed automatically). Presets create an initial frame on an unbounded document — the document itself never carries a size.
      • Canonical document creation service (createNewDocument in @varve/scene) — every creation path (home New button, empty state, File → New, Ctrl+N, command palette, template, frame preset) now funnels through one typed request that atomically produces a fully initialized document (schema version, name, optional initial frame, color config, undo state).
      • .varve native format — new saves default to filename.varve; legacy .strata documents still open through the same versioned migration pipeline. File → Save writes disk-opened documents back to their original path (Figma/Photoshop behavior). application/x-varve registered alongside the legacy application/x-strata MIME on Linux/Windows/macOS file associations.
      • Responsive workspace navigation — the editor top bar now fits the document title, workspace tabs, and controls without overlap at any width: priority-ordered workspace tabs with a "More" overflow menu, active workspace always visible, title truncating with an ellipsis.
      • Complete Varve identity — app mark now renders on the home toolbar, About dialog, custom title bar (with a graceful fallback), favicons, and the generated icon pipeline emits varve-icon.svg; remaining user-visible "Strata" strings (LUT export headers, AI diagnostics, window title defaults) renamed.

      Changed

      • The New Design dialog fits within 1280×720 viewports: sticky header and footer, internally scrollable body, viewport-capped height (also fixes the modal overlay that could intercept clicks when closed).
      • Import of native-format files preserves the original document JSON and display name instead of inserting a blank placeholder.

      Fixed

      • The closed <dialog> could remain visible and intercept pointer events (a display: flex override of the UA's dialog:not([open]) rule) — now scoped to [open].
      • Stale strata- selectors across the E2E suite refreshed to the varve- classes.
      • The colour WASM fallback referenced a build artifact that no longer exists (/wasm/strata_colour_bg.wasm → varve_colour_bg.wasm).
      • The release gate ran desktop cargo tests before the frontend existed, which tauri::generate_context!() hard-fails on — the frontend is built first now, and a workflow validator rejects any edit that moves desktop compilation ahead of it again.
      • The release draft job generated no final SHA256SUMS.txt for the merged multi-platform set — it is now generated last (over installers, manifest and SBOMs) and the draft's uploaded bytes are downloaded and re-hashed before a human can publish.
      • The SBOM generator still identified the application as Strata (tool vendor, component name, purl, one strata: property) — it now emits Varve identity and platform-scoped SBOMs with a structural validator.
      • website-deploy.yml held an unnecessary actions: write permission and had no release: published trigger, so the download page could not rebuild from a newly published release.

      Distribution hardening

      • One canonical URL system (apps/website/src/lib/siteUrl.ts): every internal link, asset, canonical URL, OG image, sitemap entry and robots location is derived from SITE_URL/SITE_BASE, so the site builds identically as the /varve GitHub Pages project site and as a root custom domain.
      • The download page is release-driven: on release: published the site is rebuilt from the exact published assets via an explicit channel policy (latest published stable, else latest published prerelease; drafts never appear) with manifest/checksum verification — an unverifiable release fails the deployment rather than inventing data.
      • Per-platform and combined CycloneDX 1.5 SBOMs ship with the release and are covered by SHA256SUMS.txt.
      • Post-deployment smoke check: homepage, download, docs, sitemap, robots, favicon, 404 and the /varve canonical prefix are verified against the live URL with bounded retries after every Pages deploy.
      • Download page accessibility: tablist semantics with arrow-key navigation, copy-to-clipboard checksums with announcements, aria-current navigation state, explicit unverified-release state.

    3. #0.1.0

      Beta
      View on GitHub

      Platform support

      Added

      • Release engineering foundation: version single-sourcing, artifact collection with predictable names, SHA-256 checksum manifests, CycloneDX SBOM generation, and a draft-then-approve release pipeline (scripts/release/, .github/workflows/release.yml).
      • Build-time guard that fails when a bundled AI model is a Git LFS pointer rather than real weights (scripts/release/check-bundled-assets.mjs).
      • Optional AI models are downloaded on demand rather than bundled, each pinned to a SHA-256 that is verified before the file is used. The installer stays around 56 MB as a result.

      Fixed

      • The release workflow could never publish: its release job depended on an AUR validation job that referenced a dist/aur directory which does not exist and is gitignored.
      • The packaged application did not reach its user interface. A static import of an eval-using module in the entry chunk stopped the frontend mounting inside the packaged WebView, and a splash screen that only closed on a signal from that frontend turned the failure into a window that could never be dismissed.

      About the name

Varve is in active development. Installers are published for Linux, macOS, and Windows.

Download Varve All releases on GitHub

Search Varve

    Browse all docs