DEV Community: Huseyn Huseynov

Building Mintu: A Privacy-First Desktop AI Companion with Tauri 2, Rust, and Local LLMs (v0.2.1 Release)

Huseyn Huseynov — Sun, 01 Nov 2026 19:49:06 +0000

Most desktop AI tools today suffer from two fundamental problems: they rely entirely on remote cloud APIs—raising serious privacy concerns—and they require constant context-switching between separate application windows.

To address both issues, I built Mintu, an open-source, local-first desktop companion that integrates directly into your screen as a native Dynamic Island. With today's v0.2.1 release, Mintu expands its platform footprint and core capabilities.

Here is a breakdown of what Mintu is, how its architecture works, and what is new in version 0.2.1.

The Core Philosophy: Complete Local Privacy

Mintu does not send prompts, conversation logs, or telemetry to external servers. All text generation and processing happen entirely on the user's hardware.

Key architectural pillars:

Zero Telemetry: No analytics beacons, tracking cookies, or external error reporters.
Local LLM Inference: Powered by llama.cpp, running quantized GGUF models (e.g., Qwen 2.5) directly on your CPU/GPU.
Local Persistence: Conversations, settings, and memory fragments are persisted strictly within a local SQLite database (mintu.db).
Offline-First: Network access is only used when the user explicitly downloads a new model or checks for GitHub releases.
The Tech Stack

Building a persistent, low-overhead desktop island that stays reactive required a stack optimized for memory efficiency and cross-platform native execution:

Core & System Integration: Rust via Tauri 2
Inference Runtime: llama.cpp bindings for quantized GGUF models
UI / Frontend: TypeScript, Vite, Modern CSS
Database: SQLite (local embedded storage)
OCR & Vision: Windows Media OCR / native image pipeline

Tauri 2 allows the application to maintain a minimal memory footprint compared to traditional Chromium-based wrappers, while Rust provides the concurrency and safety needed for background LLM inference and system monitoring.

What is New in Version 0.2.1

The v0.2.1 release focuses on workstation flexibility, productivity tooling, and multi-platform distribution:

Multi-Monitor Display Management

Users with multi-monitor workstations can now freely anchor and move the Dynamic Island across displays without layout disruption or window clipping.

Quick Scratchpad

An inline, accessible note-taking interface integrated directly into the island, allowing developers to dump code snippets, thoughts, and quick tasks without leaving their IDE or current window.

Dedicated Audio Engine

A lightweight, zero-dependency sound pipeline designed for subtle auditory feedback during focus timer transitions and background system alerts.

Native macOS Universal Package

In addition to Windows 10 and 11 (64-bit), Mintu now officially supports macOS across both Apple Silicon (M1/M2/M3/M4) and Intel architectures, integrating with Apple Keychain and native window management.

Direct Web Distribution

Users can now download both Windows installer packages (.exe) and macOS universal archives directly from the official website without relying solely on GitHub release mirrors.

Verification and Security

Because Mintu is a community-driven open-source project without commercial code-signing certificates, we publish full SHA-256 checksums for every build so users can independently verify binary integrity before execution:

Windows Installer (Mintu_0.2.1_x64-setup.exe): 670E600D8F80606C87AE949F88EB65599DFDA8825804821873F10F1D0237A5BA
macOS Universal (Mintu-v0.2.1-macOS.zip): 3C24988419EA9869AA9F76FBE2D924DA40F75C5D4F75963ED48390C2E38B5ED5
Try It Out and Contribute

Mintu is 100% free and licensed under the permissive MIT License.

Website: https://mintudev.netlify.app/
GitHub Repository: https://github.com/HuseynDeveloper/Mintu

Feedback, bug reports, and pull requests are welcomed. If you find the project interesting, consider starring the repository on GitHub.

Building Bilgo: Cross-Platform Real-Time State Sync Between Android and React Web

Huseyn Huseynov — Sun, 01 Nov 2026 14:27:50 +0000

Overview

When building real-time multiplayer web applications, supporting mobile users seamlessly is one of the most critical challenges. For Bilgo, our interactive trivia and competitive learning platform, we recently bridged the gap between our native Android app and our React + TypeScript web client.

While the Android version is currently in the Google Play review pipeline, we rolled out a standalone, signed APK distribution channel coupled with full cross-platform synchronization.

Here is a breakdown of the architectural challenges we solved and how we structured the solution.

The Challenge: Divergent Schemas Across Platforms

During the early stages of development, the Android client and the web frontend stored gameplay progress slightly differently in Cloud Firestore:

Web Client: Stored achievements as key-value maps (achievements: { [id]: { unlocked: boolean, progress: number } }).
Android Client: Persisted achievements as an array of unique identifier strings (unlockedAchievements: ["first_win", "streak_3", ...]).
Streak Telemetry: The mobile app stored streak metadata inside nested objects (streak: { count: number, best: number }), while web aggregations previously expected flat counters.

As a result, mobile players who accumulated significant streaks saw their count display as zero on the web leaderboard, and unlocked mobile medals did not register on the web profile.

The Solution: Unified Ingestion and Non-Destructive Merging

Instead of forcing a destructive migration on active player databases, we introduced a resilient aggregation pipeline on the web client:

Multi-Source Streak Extraction: A normalized extraction utility inspects all candidate structures (streak.count, streak.best, streakCount, and stats.bestStreak) and returns the true mathematical maximum across data sources.
Non-Destructive Document Merging: When fetching global and weekly leaderboards across Firestore instances, documents from publicProfiles and users collections are merged symmetrically using Math.max across scores, wins, XP, and streak counts.
Dual-Write Achievement Unlocks: When an achievement unlocks on the web, we invoke arrayUnion to append the achievement ID to unlockedAchievements while concurrently recording the structured map. This ensures future mobile sessions immediately recognize the unlocked badge.

Real-Time Download Telemetry

To measure user adoption prior to the official Google Play rollout, we implemented an in-house telemetry pipeline:

Environment Detection: Client-side detection identifies OS (Android, Windows, iOS, macOS) and device category (mobile, tablet, desktop).
Event Logging: Each APK download registers an atomic increment and writes an event log entry inside Firestore, preserving audit trails without requiring external third-party tracking scripts.
Real-Time Admin Panel: An administrative dashboard listens via onSnapshot to display real-time counters for total downloads, unique devices, and registered active mobile players.
Tech Stack Summary
Frontend: React 18, TypeScript, Tailwind CSS, Vite
Backend & Database: Google Cloud Firestore, Firebase Modular SDK (v10)
Mobile: Native Android APK distribution
Hosting: Netlify Edge
Check It Out
Web Application: https://bilgo.netlify.app
Android Download Hub: https://bilgo.netlify.app/download

Developed by the team at PixelGroup. Feedback and technical discussions are always welcome!

Bilgo

Huseyn Huseynov — Wed, 28 Oct 2026 15:25:08 +0000

Hey DEV Community! 👋

I'm excited to share a project I've been working on: Bilgo — a real-time multiplayer trivia and brain arena.

The core idea was to build a fast, competitive, and interactive quiz platform where players can challenge friends in real-time, climb competitive ranked leagues, and explore themed challenges.

It's completely cross-platform, built with Kotlin (Jetpack Compose) on Android, React on the web, and powered by Firebase for live synchronization.

I’d love for you to check it out and let me know your thoughts or feedback!

👉 Play it here: https://bilgo.netlify.app

What do you think? I'd appreciate any feedback on the gameplay, UI, or overall experience!

Building Mintu: An Open-Source Living Desktop Pet and Local AI Companion with Tauri 2, Rust, and TypeScript

Huseyn Huseynov — Sun, 18 Oct 2026 14:47:50 +0000

Most desktop AI interfaces today follow a predictable pattern: an isolated browser tab, a heavy Electron window, or a generic sidebar that gets lost in a sea of open windows.

With Mintu, I wanted to explore a different paradigm: what if your AI assistant was not an application you open and close, but an ambient, living desktop companion that resides naturally in your workflow?

Mintu is an open-source living desktop pet and local AI assistant built from scratch using Tauri 2, Rust, Vite, and TypeScript. It combines the ergonomics of an Apple-style Dynamic Island with 60 FPS kinematic pet physics, 100% private local LLM execution, real-time screen intelligence, and an open plugin ecosystem.

Architecture Overview

Building an interactive desktop pet that overlays other full-screen applications without stealing input focus or causing high CPU consumption requires strict architectural separation:

Unified Click-Through Screen Overlay Window (Tauri 2 / Win32):
Single transparent coordinate space across the primary monitor with a Dynamic Island using spring-physics morphing (width, height, radius), a kinematic desktop pet layer (gravity, momentum, window edge anchoring), and a 30 Hz Win32 cursor poller with dynamic DOM hit-rect synchronization (maintaining under 2% idle CPU).
Local AI Engine (llama.cpp Sidecar):
Automatic hardware tier detection (RAM, VRAM, CPU cores), a resumable chunk downloader with SHA-256 hash validation, real-time token streaming with syntax-highlighted markdown, and multi-provider cloud adapters (Gemini, Groq, OpenAI, Ollama).
Ambient Intelligence & Smart Clipboard:
Foreground window process inspection and security blocklist, interactive screen-region capture with offline Windows Media OCR, and a real-time clipboard classifier (code review, translation, JSON cleanup).
Extensibility & Gamification:
Focus XP progression and customizable wardrobe accessories, paired with JSON-based skin templates and runtime plugin manifests.

Key Features

1. Dynamic Island & Kinematic Pet Physics

Mintu lives inside a floating pill at the top of your screen that expands, contracts, and morphs into different states (notifications, system media, Pomodoro timer, AI chat).

When you want Mintu to explore, you can drag Mintu out or press Ctrl + Alt + M. Mintu leaves the island, sits on top of your currently active window, falls down if that window closes, and hops back into the island when thrown near the magnetic docking zone.

2. 100% Local & Privacy-First AI

Privacy is a core design principle:

Integrated llama.cpp sidecar running quantized GGUF models directly on your CPU/GPU.
No network requests are sent when using local models.
Built-in Privacy Guardian automatically suppresses context ingestion when sensitive applications (password managers, banking tools) are in the foreground.

3. Smart Clipboard

Whenever you copy text with Ctrl + C, Mintu analyzes the content in real time and offers context-sensitive actions directly in the island:

Programming code: Suggests "Check code for errors" or "Explain code".
Foreign-language text: Suggests "Translate to Azerbaijani" (or your preferred language).
JSON or unstructured data: Suggests "Format and validate JSON".

4. Multi-Language Voice & Vision

Microphone speech-to-text with automatic multi-language recognition (supporting Azerbaijani, English, Russian, Turkish, and more).
Interactive screen-region capture (Ctrl + Alt + S) that captures any rectangle on your screen and extracts text using Windows Media OCR without uploading images to any third-party server.

5. Focus Gamification & Wardrobe

As you complete Pomodoro work intervals, Mintu earns focus XP and levels up. You can customize Mintu with earned accessories from the wardrobe, including reading glasses, winter beanies, a coffee mug, and noise-cancelling headphones.

6. Developer Extensibility: Plugins & Custom Skins

Mintu v0.2.0 introduces an open architecture:

Plugins: Developers can build standalone widgets (crypto/currency rates, GitHub alerts, prayer times, weather).
Custom Skins: A transparent JSON schema allowing artists and developers to reskin Mintu with their own spritesheets and animations.

Getting Started

You can test the pre-compiled standalone binary or build from source:

Prerequisites

Windows 10 (version 1809+) or Windows 11
Microsoft Edge WebView2 Runtime (pre-installed on modern Windows)
Node.js 18+ and Rust toolchain (if building from source)

Installation

Clone the repository:


bash
git clone https://github.com/HuseynDeveloper/Mintu.git
cd Mintu
npm install