The stack. Electron, vanilla JS, no framework, no build step. Every app inside (Mail, Tasks, Goals, Portfolio...) is a plain JS object. Records live in SQLite on disk. Decision to use Vanilla JS started as an experiment to write code that will never fail, but eventually libraries were added. Anyone can download the code and run "npm start" and use the app or download the Built app from download page.
I started the project with the hope that i can build an assistant for myself in my 16 gb machine using an open weight model as the brain. but after building it, i realized even models claiming to work in 16gb m series macs does not really work that well when used as brain for an agent. As a chatbot it works. Due to this reason, I have introduced a hosted model option so people can try out the features. But it works well with models like Qwen 3.6 models in mac with more RAM.
The stack. Electron, vanilla JS, no framework, no build step. Every app inside (Mail, Tasks, Goals, Portfolio...) is a plain JS object. Records live in SQLite on disk. Decision to use Vanilla JS started as an experiment to write code that will never fail, but eventually libraries were added. Anyone can download the code and run "npm start" and use the app or download the Built app from download page.
I started the project with the hope that i can build an assistant for myself in my 16 gb machine using an open weight model as the brain. but after building it, i realized even models claiming to work in 16gb m series macs does not really work that well when used as brain for an agent. As a chatbot it works. Due to this reason, I have introduced a hosted model option so people can try out the features. But it works well with models like Qwen 3.6 models in mac with more RAM.