📐 System Architecture

Zentauri adopts a strict Rust-First and Tauri-Native approach. This guarantees maximum performance, memory safety, and future-proof compatibility with WebAssembly and advanced Rust toolchains.

---

1. Core Principles

• No Heavy Node.js Ecosystem: We avoid heavy Node.js libraries (like Playwright or Electron) for backend operations.
• Rust Backend (src-tauri): All file-system operations, workspace indexing, and data extractions happen natively in Rust.
• Dumb UI (src): The Vue 3 frontend acts strictly as a presentation layer and relies on Tauri IPC commands for data access.
• Dual Data Architecture (Plain Text & Master-JSON): Linear prose (essays, lesson notes) is canonically stored in Markdown (Mode 1). Segmented historical corpora and dictionary entries (editions like Böhtlingk or Bṛhadāraṇyaka) are canonically stored and edited as structured Master-JSON (book.json / Mode 2). External formats (TEI-P5 XML, Typst-PDF, publisher .docx) are strictly downstream compilation targets.

---

2. Component Topology

flowchart TB
    subgraph Frontend["Vue 3 Frontend (src)"]
        UI["CodeMirror 6 Editor & Vue UI"]
        MD["markdown-it-extensible (Typst AST)"]
        GRAPH["v-network-graph (Knowledge Map)"]
    end

    subgraph Backend["Rust Backend (src-tauri)"]
        IPC["Tauri IPC Command Handlers"]
        DB["rusqlite (index.db)"]
        FS["Native File System Manager"]
        TYPST["typst-as-lib (PDF Engine)"]
    end

    subgraph Storage["Workspace"]
        ZENT["/.zentauri/index.db"]
        MD_FILES["Markdown Files (*.md)"]
    end

    %% Interactions
    UI <-->|Tauri IPC| IPC
    MD <-->|Tauri IPC| IPC
    GRAPH <-->|Tauri IPC| IPC
    
    IPC --> DB
    IPC --> FS
    IPC --> TYPST
    
    DB <--> ZENT
    FS <--> MD_FILES
    TYPST -->|Write native .pdf| FS

---

3. Database & Indexing

Zentauri maintains a local SQLite database (.zentauri/index.db) within each open workspace. This allows for lightning-fast queries and powers advanced features:

• files: Tracks file paths, directory structures, timestamps, and sizes.
• markdown_metadata: Stores parsed YAML frontmatter like title, tags, iast, and devanagari values for scholarly lookup.
• markdown_links: Maps [[wiki-links]] and standard markdown links, preparing relational graph indexing (interactive UI parked on Roadmap).

WICHTIG

The indexing process leverages Rust's regex and serde_yaml crates for high-speed parsing.

---

4. Native PDF Export (Active)

Zentauri drops heavy Chromium/Playwright dependencies in favor of native Rust rendering. We integrate Typst (typst-as-lib, typst-pdf) directly inside the Tauri backend (export_pdf IPC command, ⌘P). The Vue frontend generates a Typst AST string from the Markdown document, passing it via IPC to Rust, which compiles an academic publication-grade PDF in milliseconds without external toolchains or Node.js dependencies, incorporating the STIX Two Text and Noto Serif Devanagari fontstacks.

---

5. Sanskrit Input Architecture (Hybrid Framework)

For multilingual scholarly prose (English/German + IAST/Devanāgarī), Zentauri adopts a 3-stage hybrid input architecture at the application level rather than relying on brittle OS layouts or raw ASCII Harvard-Kyoto:

1. Canonical Persistence (Active): Source Markdown files always store standard IAST (kṛṣṇaḥ) or Devanāgarī (कृष्णः).
2. In-Editor IME (Roadmap / Backlog for v2.0.0): Automatic scope-based live transliteration from Harvard-Kyoto inside 《...》 and ⟪...⟫, accompanied by compose-key sequences for running prose.
3. Post-Hoc Tooling (Active): Global shortcuts (⌥⌘D, ⌥⌘H) and toolbar toggles with tag-safe parsing for instantaneous selection toggling.

Full architectural specification and trade-off matrix: Sanskrit Input Architecture and Roadmap & Backlog.

---

6. Work-Modes Architecture (Philological Desktop Suite)

To prevent feature bloat while meeting all demands of scholarly edition projects, Zentauri structures its presentation layer into three distinct Work-Modes toggled via the Activity Bar:

1. Mode 1: Markdown Authoring (Active)
   • Single-file and split-pane text authoring (CodeMirror 6).
   • Sanskrit bracket highlighting, KaTeX formulas, and Typst AST compilation.
2. Mode 2: Synoptic Master-JSON Edition (Roadmap v2.1.0)
   • Specialized split-pane editor for historical editions (book.json / AST format as in brhadaranyaka).
   • Pairs page scans (PDF) with segmented field inputs (Devanāgarī, IAST, translation, critical apparatus) for automated compilation into TEI-P5 XML and Typst-PDF.
3. Mode 3: Dual PDF QA & Proofing (Active Core / Roadmap v2.0.0)
   • Side-by-side vector PDF comparison (hayro_svg) with synchronized scroll-lockstep, SWAP, and Finder/CLI launch.
   • Normalized bounding-box annotations, 1-click clipboard prompt for LLM proofing, and flattened annotated PDF export planned for v2.0.0.