Dou and Dillon are two in-house experimental apps I made for myself and then shared on Github as open source software. They both use Ollama or LM Studio as language model providers, meaning all the processing is done on device and privately.
They can be downloaded from my GitHub here:
https://github.com/shokuninstudio
Dou is what you get if you combine a chatbot, a Sticky notes app and a mind maps app. It’s a deceptively powerful desktop application for knowledge organisation, note-taking, and mind mapping. The name "Dou" means "way" or "path" in Japanese (Dao in Chinese and Do in Korean) reflecting the app's core idea of organising thoughts, ideas and data as text-based nodes along a path.
A path can be anything you want it to be. Your daily diet, a mood or health tracker, ideas for an app or business, a story outline, your goals ahead, or just random notes that you need help organising. You can have as many different paths you want on one canvas. Or you can create different canvases (saved as Dou files) for each path.
After you have arranged text nodes in order and linked them up, ask your favourite Ollama and LM Studio hosted large language model for feedback, summaries and more.
Dillon is a simple looking but powerful word processor designed for writers seeking enhanced productivity and organisation while being entertained. It offers comprehensive tools for efficient document creation and management and built in AI assistants who offer you feedback while you write.
Features include:
An extremely clean user interface.
Very easy to configure assistants who give feedback to you while you write. They are one click away and you can make your own just by giving it a name and character traits.
A Query window for reviewing or editing the current document with an assistant.
A Scrivener style compiler which will output to various formats such as OpenDocument, PDF and Final Draft.
Speaking of Final Draft, the Final Draft export option detects paragraph alignment, brackets and INT/EXT beginnings and converts them to the right Final Draft element.
A file format that maintains a 50 levels deep history, so if you close and reopen the app you can still undo some changes you made before you closed the document the last time.
Automatic timestamped backups.
I’ll probably be maintaining them for a while until I discover a few bugs and after that they will be feature locked. As long as they do the job they are supposed to do (help me process my thoughts…yes, you can use a language model to help process your own thoughts, ideas and feelings) then there won’t be any need to add more bells and whistles.
I am currently working on a suite of apps for animators. They do not use "generative AI”, but I am attempting to do things nobody else is doing and focusing on efficiency and performance. The goal is to make animation easier, faster and less stressful simply by creating clean interfaces, fast user interaction and smart algorithms.
Like Dou and Dillon I am building them primarily for myself because so many animation apps have become bloated with far too many features that slow down performance on even the fastest computers. If I find my ideas work really well, I’ll share the apps freely with the world - some open source and some not.