Show Vue: Knovum – A local-first desktop AI workspace built with Vue 3 (Composition API, Pinia, Konva & Tauri 2)

[madebyzan]

Hi Vue community! 👋

I wanted to share Knovum, an open-source, local-first desktop AI knowledge workspace built using Vue 3 Composition API, Pinia, and Tauri 2.

We wanted to push the boundaries of what is possible with Vue 3 in a native desktop environment without falling back to heavy Chromium wrappers (Electron).

🛠️ How we use Vue 3 in Knovum

1. Reactive 2D Canvas Engine (Konva + Vue 3):

   • Infinite 2D workspace rendering notes, PDF excerpts, audio cards, and relation arrows.
   • Spatial node geometry is managed through isolated reactive refs with viewport culling, keeping canvas rendering smooth at 60fps even with hundreds of nodes.
   • We maintain a strict separation between high-frequency desktop IPC events and Vue's reactive store to prevent micro-stutters during panning.

2. Clean Component Architecture & Strict Gate Checks:

   • Built entirely with <script setup lang="ts">.
   • Capped TS/Vue file lengths at 180 lines and cognitive complexity at 10 to ensure maintainability.
   • Feature-driven modular layout where each capability owns its UI and state.

3. Desktop Native & 100% Offline (Tauri 2 + Rust Core):

   • Built with Tauri 2, SQLite, and embedded LanceDB vector search.
   • Idle memory footprint is only ~45MB (compared to 800MB+ in typical Electron apps).
   • Natively supports the open JSON Canvas (.canvas) standard, allowing seamless interoperability with Obsidian.

---

• GitHub Repository: https://github.com/madebyzan/knovum
• Tech Stack: Vue 3, TypeScript, Pinia, Tailwind CSS v4, Reka UI, Konva, Tauri 2 (Rust)

We'd love to hear feedback and thoughts from the Vue community on state management patterns for desktop apps and high-performance canvas integrations!

[madebyzan]

🚀 Update: Knovum v0.0.2 released with Streaming AST Block Boundaries & Auditionable Voice Engine

We just shipped v0.0.2 with key improvements to our Vue 3 frontend architecture:

• Streaming Conversation Block Boundaries: Streaming assistant responses are now parsed into discrete AST blocks on-the-fly with single-writer ownership in Vue Composition API. This prevents full DOM re-renders and eliminate visual layout shifts during high-throughput token generation.
• Auditionable Voice Library: Added an interactive voice preview library with dynamic speaker waveforms built with Vue reactive primitives.
• Differential Space Indexing & Task Center Bridge: Wired the desktop Task Center to Tauri Channels with live cancellation tokens, allowing long-running LanceDB re-indexing to stream progress to the Vue UI without locking the event loop.
• Tailwind v4 Semantic Token Refresh: Hardened dark-mode canvas contrast and unified overlay styling.

Pre-built binaries (macOS DMG & Windows installer) and full changelog:
https://github.com/madebyzan/knovum/releases/tag/v0.0.2