DOST: Your Companion

• 

  Home

• 

  Connection Session with DOST

• 

  DOST ready to hatch!

• 

  DOST Notification

• 

  Chill DOST

• 

  DOST Home

Inspiration

I’m a software engineer, so most of my day is spent at my desk. I noticed this lifestyle was starting to hurt me, but I still kept forgetting to take small breaks. So I made a companion to remind me.

What it does

DOST means “Friend” in Urdu. It’s a small desktop companion that lives in your Windows tray.

Seed phase: DOST starts as an egg. Over three short incubation sessions, it asks you five light questions each time to get to know you. Your answers decide which persona hatches: Chill (gentle, green) or Bossy (playful, orange-red).

Personal messages: When it hatches, a local AI model uses your answers to write DOST’s nudges, wisdom and jokes, so they feel like they’re meant for you.

Gentle nudges: Every couple of hours, DOST reminds you about the habits you said you wanted to build. In between, it sends light encouragement and jokes. Reminders are small pop-ups that don’t block your work, and you can mute them when you need to focus.

Care for it: You can Feed DOST or put it to Sleep, which also mutes reminders for five hours.

How we built it

I used the skills provided in this hackathon to find and shape the idea. Before starting it I had no project idea I just believed what was written .that it would help me find one and Yes it did!

Invoking skills one by one: I started with building scope with ChatGPT. Then I proceeded to build the product requirements and finally built the technical specifications.

After these steps I started implementation. I implemented it with Claude Code and ChatGPT. The scope, PRD, and technical specs finalization made it easier for AI to implement major functionalities.

Another best direction by developer skills pack was to first make a POC. That the main functionality is possible or not. So I built the main functionality with dummy data first and confirmed each functionality was possible and working. Then proceeded to properly finish an end to end working application.

Developer skills helped me make this app how apps must be built "block by block" and then connect the blocks.

Technology Used

1. Electron + Node.js for the Windows desktop app, tray and reminder pop-ups
2. Ollama running Qwen 3.5 (4B) locally to write the personal messages Everything (your answers and messages) is saved only on your PC

Challenges we ran into

Choosing an LLM was the hardest decision. A cloud LLM would mean sending details of the user’s everyday life to someone else’s servers. I decided to use a local model instead, so the user’s information never leaves their computer.

That brought its own challenges. A small local model is slower and less predictable, so DOST writes all its messages once at hatch time instead of on every reminder. If Ollama isn’t installed or generation fails, DOST still hatches with its own prepared messages, so the app always works.

Accomplishments that we're proud of

Seeing the whole idea come together, step by step, was really encouraging. I’m also proud that DOST is fully private: it gives personal, AI-written encouragement without any of your data going to the cloud. Cherry on top is DOST's persona. Its not generic but it can be a companion with a personality.

What we learned

AI is an extremely useful tool for bringing ideas to life if you use it correctly and in the right order. Defining the scope, and being reminded whenever I was growing it, kept everything on track. The biggest lesson was to build step by step: plan first, then requirements, then the technical implementation, instead of diving straight into code.

What's next for DOST: Your Companion

DOST has a lot of potential to become a real companion that helps with communication and handles some of your tasks. The part I most want to grow is DOST’s own personality: more personas, and a companion that keeps learning about you over time instead of only during the seed phase.

Built With

• electron
• node.js