June 4, 2026

Building the Ultimate Private AI Task Manager: My Self-Hosted “Todoist” Upgrade

In the world of productivity, Todoist is often hailed as the gold standard for its natural language processing. But as someone who values data privacy and the flexibility of open-source software, I wanted something more: a system I own, that lives on my hardware, but doesn’t sacrifice that “smart” AI feeling.

Today, I’m excited to share how I built a seamless, AI-powered task management workflow by connecting three powerful tools: VikunjaFlow Launcher, and OpenRouter (Gemini AI).

The Stack: Why These Three?

To create a truly “seamless” experience, I needed three distinct layers to work in perfect harmony:

  1. The Database (Vikunja): My self-hosted, open-source replacement for Todoist. It’s powerful, fast, and most importantly, the data stays with me.
  2. The Interface (Flow Launcher): I don’t want to open a browser every time I have an idea. Flow Launcher allows me to hit a hotkey and type a task instantly from anywhere in Windows.
  3. The Brain (OpenRouter/Gemini): This is the “brain” that connects the two. By using an AI API, I can turn messy, natural language notes into structured, categorized tasks.

Adding the “AI Flavor”

While Vikunja is great out of the box, I wanted to add a layer of intelligence that felt like magic. By developing a custom Python plugin, I’ve achieved a level of automation that even most paid apps don’t offer:

1. Natural Language, Decoded

I can type 

task buy milk tomorrow at 5pm

 or 

task gym every monday

. The AI understands the relative dates, the specific times, and even handles complex scenarios like Year Rollover. If I type “New Year’s visit on 1jan” in late December, the AI is smart enough to know I mean next year, not the one that’s almost over.

2. The “Fun” Factor: Auto-Emoji

Productivity shouldn’t be boring. The plugin analyzes the task and prepends a relevant emoji.

  • “Buy milk” becomes 🥛 Buy Milk
  • “Gym” becomes 🏋️ Gym
  • “Museum visit” becomes 🏛️ Museum Visit

3. Intelligent Auto-Labeling

This is where it gets really smart. Instead of me manually tagging every task, the AI looks at my existing Vikunja labels.

  • If I have a “Shopping” label, it reuses it for groceries.
  • If I’m adding something entirely new, like “Fix the car,” it can auto-create a “Maintenance” label on the fly.

The Workflow: From Thought to Task in 2 Seconds

The beauty of this setup is the friction-less experience:

  1. F1: Open Flow Launcher. (I chose F1 specially because that’s the nearest key which remain unused till this date)
  2. Typetask meeting with the team next Friday at 10am.
  3. Enter: The AI parses the date, adds a 🤝 emoji, classifies it under “Work,” and sends it straight to my self-hosted Vikunja instance.

Why This Matters

By building this, I’ve achieved a True Fully Controlled Task Manager. I’m not paying (planning to not pay) a monthly subscription, my data isn’t being mined for ads, and I have a system that is custom-tailored to exactly how I work.

It’s a reminder that with the right APIs and a bit of “glue” code, we can build tools that are more powerful than the biggest names in Silicon Valley, all while keeping our digital sovereignty intact.

All made possible via Google Antigravity tool.

0 0 votes
Article Rating
Subscribe
Notify of
0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x
WordPress Appliance - Powered by TurnKey Linux