AI Agent - Deep Search & Note Summarization

Knowledge Graph - Visualize Note Connections

Editor - LaTeX Formulas & Backlinks

---

Lumina Note is not just a Markdown editor – it is an LLM-native knowledge workspace:

• Local Markdown vault, Git-friendly
• Built-in multi-provider LLM client (8 providers)
• Full-featured Agent system with tools
• RAG semantic search on your own notes
• PDF reader + annotation system
• Dataview-style databases driven by YAML
• Bilibili video notes, voice notes, daily notes

---

• Three editor modes: Source / Live Preview / Reading
• WikiLinks: [[WikiLinks]] to build a graph-like knowledge network, supports drag-and-drop from file tree
• Beautiful formatting: LaTeX math (KaTeX), Mermaid, Obsidian-style callouts > [!info]
• Syntax highlighting: CodeMirror 6 with hundreds of languages
• Split view editor:
  • Horizontal / vertical split
  • Draggable divider with live resize
  • Active pane tracking: the pane you clicked last becomes the target when opening files
• Image paste: Ctrl+V to paste screenshots directly into editor, auto-saved to vault

• Canvas-based, high-performance visualization of your vault
• Folders shown as spiky balls, children inherit folder color
• [[WikiLinks]] automatically become edges between notes
• Right-click a node to open an isolated view (node + neighbors)
• Physics simulation for natural clustering & interactive dragging/zooming

• Multi-provider LLM support:
  • Anthropic, OpenAI, Gemini, DeepSeek, Moonshot (Kimi), Groq, OpenRouter, Ollama (local)
• Agent modes: editor, organizer, researcher, writer
• Tooling: read/edit/create/move/search notes, database tools, RAG search, grep, etc.
• Real-time edit preview: animated diff playback inside the editor
• RAG semantic search: vector DB + optional reranker
• AI floating ball: draggable assistant that doesn’t break your flow
• Voice input: streaming STT with auto-stop and animation

• Play Bilibili videos inside the app (Tauri multi-WebView)
• Send special-prefixed danmaku to create timestamped notes
• Click timestamps in notes to jump the video
• Notes are saved as Markdown and reloaded automatically next time

Optimized for academic and research workflows.

• Interactive element detection: detect text, images, tables, etc.
• Annotation system:
  • Highlight / underline selected text (5 colors)
  • Add textual notes to highlights
• Annotation storage:
  • Saved as pure Markdown files: yourfile.pdf.annotations.md
  • Lives next to the PDF, Git-friendly and editable in any editor
• Bi-directional jumping:
  • From PDF → annotation file
  • From annotation file → back to the exact PDF location via lumina://pdf links
  • Ctrl+Click opens the PDF in split view and jumps to the annotation
• Thumbnails & outline: page thumbnails and table-of-contents sidebar
• Full-text search: in-document search with highlight
• AI integration: send selected PDF content to AI for summarization/translation

• 15 built-in themes, each with light + dark variants (30 looks in total)
• Themes affect headings/links/code/blockquote and 17+ Markdown elements
• Custom title bar that follows theme colors, with window controls
• All theme settings are managed in the Settings panel (bottom-left gear icon)

• YAML frontmatter-driven: notes are the source of truth
• Table and Kanban views
• 7 column types: text, number, select, multi-select, date, checkbox, URL
• Database definitions stored as .db.json (structure only, no row data)
• File tree integration: database icons, click to open views
• Fully Git-friendly and plaintext-friendly

---

• Node.js ≥ 18
• Rust ≥ 1.70
• npm or pnpm

1. Clone the repo

git clone https://github.com/blueberrycongee/Lumina-Note.git
cd Lumina-Note

2. Install dependencies

npm install
# or
# pnpm install

3. Run in dev mode

npm run tauri dev

4. Build production app

npm run tauri build

To enable advanced interactive PDF element selection:

1. Install Python deps (first time only):

cd scripts
pip install flask flask-cors pymupdf

2. Start the server:

python simple_pdf_server.py

The server runs at http://localhost:8080.

Note: In the long term this Python service will be migrated to a Rust Tauri command so no extra runtime is required.

---

• On first launch you will be asked to select a folder as your vault.
• ⚠️ Important: do not choose huge directories (like your entire Documents or home folder).
  • Too many files will slow down indexing and RAG semantic search.
  • Recommended: create a dedicated folder like ~/LuminaVault or ~/Notes.

• Create a note: sidebar + button or Ctrl+N.
• Switch editor mode (buttons in the top-right of editor):
  • Source – raw Markdown
  • Live Preview – rendered view while typing (recommended)
  • Reading – clean reading view, no editing
• WikiLinks: type [[ to link to other notes (with auto-complete).
• Tags: use #tag anywhere in the note; tags are indexed in the right sidebar.

• Click the calendar icon in the sidebar.
• Creates a new Markdown file named like Quick_2025-12-02_06-00.md.
• Perfect for meeting notes, daily logs, and scratch ideas.

• Click the microphone icon in the sidebar.
• Speech is transcribed live into the editor.
• Recording auto-stops after a few seconds of silence.
• Button shows an animated ripple while recording.
• Implemented with the browser’s Web Speech API – no extra install needed.

• Open via the sidebar graph icon or command palette.
• Interactions:
  • Click a node to open its note.
  • Drag nodes to reposition them.
  • Right-click a node to open an isolated view (only that node + neighbors).
  • Scroll to zoom in/out.
• Visual features:
  • Folders as spiky balls, auto-colored per folder.
  • Child notes inherit folder color.
  • [[WikiLinks]] become edges between notes.

• Click the gear icon in the bottom-left to open Settings.
• Choose from 15 built-in themes, each with light & dark mode.
• Quickly toggle light/dark via the sun/moon icon in the title bar.

1. Configure API keys:

   • Open Settings → AI Settings (bottom-left gear icon)
   • Choose a provider and fill in your API key / base URL if needed
   • Click Save

2. Where you can use AI (all support Chat & Agent modes):

   • Right-side AI panel
     Use the AI tab in the right sidebar to chat or run agents with the current note as context.
   • Floating AI ball
     Enable in Settings. A draggable floating button appears; click it to open Chat / Agent anywhere without leaving your current view.
   • Main-view AI mode (left ribbon button)
     Click the AI icon in the left ribbon to open a full-screen AI view in the main area – ideal for long conversations or complex Agent workflows.

3. RAG semantic search:

   • Turn on RAG indexing in Settings.
   • The Agent can then semantically search your whole vault to answer questions.
   • Note: the first indexing pass can be slow on very large vaults, so keep your workspace reasonably sized.

1. Open a PDF from the sidebar.
2. Select text to bring up the annotation toolbar.
3. Choose highlight color / underline or add a text note.
4. An annotation Markdown file <pdfname>.annotations.md is created/updated automatically.
5. In the annotation file, use the [📍 Jump] links:
   • Normal click: open PDF in the main pane and jump.
   • Ctrl+Click: open PDF in the split pane and jump.

• Toggle split view via the editor toolbar.
• Click left/right pane to mark it as active (highlighted border).
• When split view is on, opening a file from the sidebar will open it in the active pane.
• Drag the divider to resize panes.

• Define database structure in a .db.json file.
• Attach notes to a database via YAML frontmatter (db: projects).
• Use table or Kanban view to browse & edit; changes sync back into note YAML.

---

See the Chinese README or docs/ for more implementation details. In short:

• Frontend: React 18 + TypeScript + Tailwind + Zustand + CodeMirror 6
• Backend: Tauri v2 + Rust commands for filesystem and vector DB
• RAG: SQLite-based vector store, multiple embedding backends, optional reranker

---

• Core: Multi-provider LLM support (8 providers)
• Agent: AI Agent system with tool calling (4 modes + 14 tools)
• RAG: Local vector database with semantic search (+ Reranker)
• UI: AI floating ball with streaming responses
• Video: Bilibili video notes (embedded WebView + danmaku sync)
• STT: Voice-to-text (streaming + auto-stop + recording animation)
• Theme: 15 built-in themes + Markdown color customization
• Settings: Central modal settings panel
• PDF: Smart PDF reader (element detection + interactive selection + AI chat)
• PDF Annotations: Highlight/underline/notes, Markdown storage, bi-directional jump
• Split Editor: Resizable split panes with active panel tracking
• Database: Dataview-style YAML-driven table/kanban views
• Image Paste: Paste screenshots directly into editor, auto-save to vault
• Drag-and-drop WikiLinks: Drag files from tree to editor to insert [[WikiLink]]
• WebDAV Sync: Support Jianguoyun and other WebDAV services for bi-directional vault sync
• Flashcard System: Generate flashcards from notes, spaced repetition review (basic/cloze/MCQ)
• Multi-language: Chinese/English/Traditional Chinese/Japanese internationalization
• Dynamic Routing: Intent recognition + multi-model routing for smart task distribution
• Backlinks Panel: Show which notes reference the current note, click to navigate
• Tag Aggregation: Tag list with counts, expand to see related notes
• Export PDF: One-click export notes to PDF files

• PDF Rust Migration: Port Python PDF parsing service to Rust
• Calendar View: Calendar view for databases

• Template System: Choose templates when creating new notes
• Export HTML/Word: More export formats
• Plugin System: Plugin API design
• Git Sync: Git version control integration
• Mobile: iOS / Android adaptation

---

---

Lumina Note is open-sourced under the Apache 2.0 License.