HelixCanvas

Open-source, local-first biomedical illustration studio for publication-ready research figures.

Open libraries, provenance-first imports, optional AI planning and image generation, and export-ready composition in one workspace.

HelixCanvas In 30 Seconds

HelixCanvas is a local-first biomedical figure editor built for people who want publication-ready visuals without losing provenance, attribution, or control. It combines:

• a searchable open illustration library based on Bioicons
• curated Servier Medical Art assets and official kit links
• a safe import lane for user-owned FigureLabs exports
• a serious figure editor with layouts, exports, snapshots, and review notes
• an optional server-side AI copilot for planning, critique, editable scene edits, and OpenAI Image 2 content generation

The project is being shaped as a public-good tool rather than a commercial SaaS product. The long-term ambition is closer to an Inkscape-for-biomedical-figures than a locked-down hosted platform.

What Ships Today

• Source-aware built-in asset packs with pack validation, license strategy, and provenance metadata
• Bioicons search plus Servier-derived vectors, Servier originals, and safe user-owned imports
• Real biology example figures and tutorial artifacts for signaling, CRISPR workflow, and retinal degeneration
• Guided Figure Flows for signaling, methods workflows, and microscopy comparisons with structured prompts and AI handoff
• Starter-template gallery for oncology, immunology, microscopy, neuroscience, infection, and neutral workflow boards
• Insertable scientific builders for membrane pathways, compartment maps, assay strips, and timecourse timelines, now with domain-aware variants and style presets
• One-click domain presets that combine guided flows, scientific scaffolds, builder variants, and figure themes
• Figure themes plus related-search suggestions for faster refinement once a project is in motion
• Export targets for manuscript, slides, and poster work plus downloadable review bundles
• Multi-select, marquee selection, grouping, transform handles, rotation, resizing, flipping, match-size, fit-to-board, lock, and hide controls
• Layout intelligence for selected objects: align/distribute, tidy row, tidy column, tidy grid, radial ring, and panel-placement helpers
• Smart connectors with endpoint anchoring, activation/inhibition/link semantics, straight/elbow/curved routes, dashed/dotted styles, and on-canvas curve bending
• Fragment and style clipboards for copying editable figure chunks or visual styling without flattening the scene
• Layer ordering with top-first layer panel, per-layer up/down controls, one-step ordering, and front/back jumps
• Asset crop, mask, fit, zoom, opacity, flip, and visual-effect controls with export fidelity
• Panel-layout presets, legend blocks, callout blocks, scale bars, and reusable components
• Local project open/save, recovery drafts, named local snapshots, snapshot compare, and figure branching
• Inline pinned review comments that stay in the project but stay out of exports
• Semantic asset retrieval plus domain starter kits for oncology, immunology, neuroscience, and microscopy workflows
• SVG, PNG, PDF, JSON, and citation-bundle export paths
• Optional AI brief-to-plan drafting, edit-by-instruction, and figure critique with the API key kept on the server
• Optional OpenAI Image 2 generation that creates provenance-tagged, user-generated bitmap assets and places them on the canvas
• Command palette workflow for fast local actions and AI-powered figure edits
• Installable offline-ready app shell with service-worker caching for local-first use
• Desktop packaging for macOS with a branded app icon plus native open/save dialogs

Why It Feels Different

• Local-first: the core editor is useful with no account and no cloud dependency.
• Provenance-first: assets keep source and license context instead of becoming anonymous clip art.
• AI-optional: the editor works without AI, and the AI path is structured rather than magical.
• Contributor-friendly: packs, examples, docs, and validations are meant to be extendable in the open.

Project Docs

• Editor guide
• Open-source roadmap
• Personal workbench direction
• GitHub milestone plan
• Product overview
• Asset pack spec
• Real biology tutorial
• Contributing guide
• Code of conduct

Start Locally

1. Clone Bioicons locally:

git clone --depth 1 https://github.com/duerrsimon/bioicons /tmp/bioicons

2. Build the library manifest:

npm run build:library

If you have a local Bioicons checkout, point the script at it with BIOICONS_DIR=/path/to/bioicons. If not, the script will reuse the checked-in Bioicons cache and still regenerate the pack manifest.

3. Install dependencies:

npm install

4. Start the app plus local API:

npm run dev

Optional desktop build on macOS:

npm run build:desktop-icon
npm run desktop:dir

That produces a packaged app bundle in dist-desktop/mac-arm64/HelixCanvas.app with native project-file open/save dialogs.

5. Optionally configure AI:

export OPENAI_API_KEY=your_key_here
export HELIXCANVAS_OPENAI_IMAGE_MODEL=gpt-image-2

Use .env.example as the reference for local configuration. HELIXCANVAS_OPENAI_IMAGE_MODEL defaults to gpt-image-2; generated images are stored as user-generated local assets with prompt/model metadata instead of being folded into the bundled open libraries. If you skip AI setup, the editor still works and the AI controls simply stay optional.

Quality Checks

Run the local verification suite with:

npm run check

Useful supporting commands:

npm run check:packs
npm run build:library
npm run build:tutorial

GitHub Actions runs the same checks on pushes to main and on pull requests through .github/workflows/ci.yml.

Current Architecture

Editor

• src/App.jsx — main editor experience and local-first workflow
• src/lib/editorSelection.js — selection, marquee, transform, layout, layer-order, alignment, and guide helpers
• src/lib/connectors.js — connector geometry, anchoring, curve bending, dash styles, and smart connector drafts
• src/lib/sceneClipboard.js — editable fragment copy/cut/paste and graph-safe deletion helpers
• src/lib/styleClipboard.js — layer and connector style transfer helpers
• src/lib/domainPresets.js — one-click preset recipes that combine flows, themes, and scaffolds
• src/data/starterTemplates.js — curated starter-template entry points for concrete figure archetypes
• src/lib/figureFlows.js — guided figure-builder definitions and scene-graph generation
• src/lib/figureThemes.js — palette systems and current-figure restyling helpers
• src/lib/scientificBuilders.js — insertable scientific scaffold blocks with configurable variants and style presets
• src/lib/exportPresets.js — manuscript, slides, and poster export target helpers
• src/lib/layoutPresets.js — manuscript panel layout logic
• src/lib/projectFiles.js — local project file helpers and validation
• src/lib/projectSnapshots.js — named local checkpoints
• src/lib/projectCompare.js — snapshot-to-current comparison summaries
• src/lib/reusableComponents.js — reusable motif and component storage
• src/lib/reviewComments.js — pinned review notes that persist locally
• src/lib/exporters.js — SVG, PNG, PDF, JSON, and attribution export helpers

Assets

• src/lib/assets.js — ranking, semantic retrieval, recents, favorites, and suggestion helpers
• src/lib/assetPacks.js — pack schema, normalization, validation, and summary helpers
• src/data/domainStarterKits.js — domain-tuned starting points for common biological figure workflows
• packs/ — first-class committed asset pack files for built-ins and examples
• public/packs/ — local assets referenced by committed pack files
• scripts/validate-asset-pack.mjs — pack validation CLI used locally and in CI
• scripts/generate-bioicons-index.mjs — Bioicons indexing pipeline
• public/data/library.packs.json — generated built-in asset pack manifest

AI

• src/lib/ai.js — browser client for local AI endpoints
• server/index.mjs — local API server and production host
• server/aiService.mjs — OpenAI orchestration, structured output contracts, and Image API generation

Desktop

• desktop/main.cjs — Electron shell, embedded local server startup, and native file dialogs
• desktop/preload.cjs — safe desktop bridge for project-file open/save
• scripts/build-desktop-icon.sh — macOS icon generation from the shared SVG brand asset

Roadmap Right Now

The repo has already moved beyond “blank-canvas prototype” territory. The active push now is to finish the highest-leverage parity work while keeping the open-source differentiators strong.

Current focus:

• richer text controls for publication-specific labels, symbols, and inline scientific notation
• deeper connector labeling and pathway semantics beyond the current activation/inhibition/link model
• stronger retrieval, starter kits, and pack discovery across more biological domains
• more tutorials and example projects that exercise the new editor power tools
• contributor-facing architecture docs for editor, exports, assets, and optional AI boundaries

See docs/OSS_ROADMAP.md and docs/GITHUB_MILESTONES.md for the concrete milestone structure.

Community

Good contribution areas right now:

• editor polish and interaction design
• asset metadata and pack structure
• export fidelity and citation workflows
• documentation and tutorials
• accessibility improvements
• optional AI provider integrations

Sources

• Bioicons
• Bioicons GitHub
• Servier Medical Art
• Servier Image Kits
• FigureLabs

Notes

• Bioicons licenses vary by asset and are preserved in the generated manifest.
• Servier Medical Art content requires attribution and is surfaced with compliance guidance.
• FigureLabs is treated as an import lane rather than a bundled stock corpus.
• Review comments and local snapshots stay in project workflows but do not pollute exports.
• HelixCanvas is released under the Apache-2.0 license.