Building a Minimalist, Private AI Workspace
In an era where AI interfaces are becoming increasingly cluttered with subscription prompts, social sharing buttons, and tracking pixels, I wanted something different. I wanted an AI assistant that felt like a native extension of my workspace: fast, private, and distraction-free.
That’s how AI Workspace was born.
The Philosophy
The core design principle for this project is simple: Less is more.
I stripped away every unnecessary element. There are no marketing banners, no complex sidebar navigations, and no unnecessary UI bloat. The aesthetic is strictly Apple-inspired—deep blacks, subtle grays, and frosted glass (backdrop-filter) effects that respect the content above all else.
The Architecture
The system is built entirely on the Cloudflare serverless ecosystem. It is a full-stack application that proves you don’t need a heavy infrastructure to run a high-performance AI tool.
1. The Backend (Cloudflare Workers)
The entire application logic, from the API routing to the chat handling, lives in a single, optimized JavaScript worker. It handles requests in milliseconds, keeping the latency invisible.
2. The Memory (Cloudflare D1)
To ensure the AI feels “personal,” I integrated Cloudflare D1. It functions as a serverless SQLite database, allowing the system to remember conversation context. Whether I am on my PC or my phone, my workspace is synced.
3. The Brain (Workers AI)
I chose the Qwen 3 (30B) model, hosted natively on Cloudflare’s global edge. This allows for rapid, logic-heavy responses without the complexity of managing third-party API keys or worrying about unpredictable bills.
Privacy by Design
Privacy isn’t just a feature; it’s the foundation.
- Guest Mode: If I just need a quick answer and don’t want to leave a trace, the system defaults to “Guest Mode.” Sessions are ephemeral and vanish the moment the browser tab is closed.
- Sync Mode: For deeper work that requires history, a secondary encrypted login layer ensures that my chat logs are stored securely in my own database, accessible only by me.
Experience It
The project is live and functioning as my daily driver for coding tasks, logic refinement, and brainstorming.
You can visit the live workspace here: https://chat.dalei.uk
Final Thoughts
This project has been an exercise in restraint. By limiting the technology stack to a single, serverless provider and focusing on core usability, I’ve managed to create a tool that is more useful than many of the bloated alternatives available today.
The future of software is in the hands of those who know what to remove, not just what to add.
