Livygo is an application to let you customize... literally everything while keeping Duolingo gamification factors. This Project also being coded completely using DeepSeek R1 and voiced using AI TTS, while keeping its gate wide open to those who want to create their own courses ( Community Courses ) or improve the codebase. This is nothing much really when you have Anki and Lingonaut Project.

Alright I am back. During this 3 months, I have been laying off this project to focus on something else in my personal life, and will probably return to focus after a couple weeks. I am actually kinda failed to reach the goal I have set in March partially due to me procrastinating this project. But welp, this supposed to just be a fun project so it is fine, I can add stuff when I have some brilliant idea or something anytime. And no, I didn't wish to abandon this project anytime soon even though it might be in a hiatus.

Anyway off from this personal update, in this Livygo update :

1. Fix the unresponsive state when try to enter a lesson. I have aware of this and then forgot to give this repo an update when I fix it.
2. Add New Course page is added, although for now it is just a decoration, but in that page it show courses that I planned to create in the future.
3. Revamp Japanese for English speakers course into much complete sample course. This is not supposed to be a "from-scratch-to-fluency" course.
4. You can now access to the editor folder to open up a GUI to edit course (Circles, Pathway Structure, Lessons). There will be some settings missing like the inability to toggle allowUnlimitedPracticePool and the inability to create Type 6 questions through it,
5. Spanish for English speakers course's Guidebooks have edited although it is incomplete.
6. Fix Distractors only focusing on current Circle issue.

1. Introducing Question Type 5, a drag-and-drop one-to-many-subquestions question. Basically you got Word Blocks and you drag it to the respective blanks space. You can also type your answer in Type Answer mode.
2. Add Relearn Pool/New Words indicator. This is for future further development of question shuffle algorithm. It will show up either Green or Yellow on top left corner of your browser. For the time being, it's limiting to only Type 1 and Type 2.
3. Revamp the question shuffle algorithm completely from being randomized to randomize according to Question Types Ratio and New Question Ratio.
4. Adding default profile. This is just a decoration for Offline Mode and you could change whatever profile you want in profile.json and /images/. Obviously this will be different once it is online.

There are still minor issue within this update. As I suspect that the relearnpool and practicecirc in lesson_progress.json toggle simply do not work. Setting relearnpool off doesn't turn off the utilization of Relearn Pool and setting the practicecirc to true doesn't actually forcefully disable New Words Pool from being utilized. Also multiple settings that are unused ( Allowed Variants in that OCR settings for example ) should have remove from this update but I forgot. Eh, I will fix this after couple of more updates.

Next update would be focusing on course questions. This gonna be a hell of a workload to work with so wish me luck.

This is a much smaller update as I have difficulty figuring up new Question Type. In spite of that, it have arrive and working OK now.

1. License Change as mentioned above.
2. Introducing Question Type 6, Type 6c to be precise for Japanese Kanji ( it will rename to 6a once I merge it with Hanzi and Hanja ). This is a character-stroke recognition question that only appear in Handwrite Circle (handwritecirc in lesson_progress.json). Basically what you do is write on a canvas and the code will check whether it is in correct order or not.
   • The process of develop this have throw me out. I originally want to utilize KanjiVG to do some kind of SVG-stroke check, however it ended up a catastrophe with the entire expected stroke inverted, and that the code have no clue how to deal with startPoint and endPoint.
   • Even when I change the way the codebase to recognize stroke using 6-points circle measure. It works surprisingly OK, but it might have issue expanding to much more specific stroke. (CURVE is stated in the code but I have no idea what is the stroke for the code to recognize CURVE without getting LEFT_FALLING or RIGHT_FALLING)
   • OCR Validation is not available to make it more local-available.
3. Remove clutter in question.html. Also met the same fate as last update's mainnightly.html.

Yup, that's it. Just three. The next step would be Type 5 ( many-subquestion phrase/grammar/vocabulary question ) and Type 1a 1b ( Split translation and word for either answer or question ).

Go to the 4th Circle in JP-en General course if you want to try Type 6 questions.

Demonstration Video

Former changelog have been archived in Pastebin. Take a look if you are interested in development progress.

git clone this repository, and make sure you have Python 3 in order to run this web. In Command Prompt, type the following command.

python -m http.server 7500

After that, go to your web browser and enter http://localhost:7500/mainnightly.html.

If you wish to open the Editor GUI, enter this command in Terminal/Command Prompt (which you open it in the editor folder)

python courseeditor.py

• Create a Official Course Pathway

  • Spanish (EN)
    • VOCABULARY
    • GRAMMAR
  • Japanese (EN)
    • GENERAL
  • English (CN)
  • Malay (CN)
  • American Sign Language
  • Chemistry Principles (NON-LANGUAGE)

• Finish UI Decorations

• Write a more completed Guidebooks

• Hookup AI TTS Voices on Courses

• Adaptability of Community Courses

• Polish the Logo

• Possible Flashcard Mode and Anki Import (Erm, I guess?)

• Livy on Main Page

• Alpha Web Launch (not so soon)

• Android/IOS Version

• DeepSeek R1 by DeepSeek

• Gemini 2.5 Pro by Google

• GPT o3-mini and 4o by OpenAI (Served as Alternative for DeepSeek)

• GPT SoVITS by RVC-Boss

• Fish Speech S1 and S1 mini by Fish Audio

• RVC by RVC Project Team

• WanaKana by WaniKani

Well, if you are here to blame my blatant usage of LLMs or AI, sure go on. I do not code my own code.

This doesn't mean I am illiterate with code syntax and content however. In regards to literacy, there are real vulnerabilities when you entrust LLM on cybersecurity/API stuff, especially when YOU simply vibe coding and never ever understand even a slightest of their code. Even though it delivers, you as a developer is putting yourself in a great risk of getting DDoS attack once you put yourself publicly. Not like you should avoid entrust LLM for anything, but remember I hold the responsibility to my Internet accounts, and I do not endorse any idea to allow my code to attack my networking for the slightest.

Hence, this repository will be ENTIRELY local-based and offline-available. Stuffs like Profile and Shop will be set using a default JSON which you can edit whatever you like. ( Although it's decorative with no real purposes offline ). I still planned to launch an online version of the application ( with some alteration of the code ) but backend development is a field that I have never enter before.

This Project use GNU GPL v3 as its license. This license allows you to distribute freely and use for COMMERCIAL PURPOSES given you as the user have credit this repository. Modifications of this repository should have the SAME LICENSE as this repository.