SpeakCuba
Veteran
I don’t know if you’ve mentioned it before, but what language are you planning on using to program this? Apologies in advance if you have mentioned this before.
Gecko777
Veteran
i love tinkering so good luck!
F'prime
Veteran
print("I will be using Python")
ProStar
Legend
Boo no semicolons
F'prime
Veteran
Attachments
ProStar
Legend
I like not having to indent properly
F'prime
Veteran
THE GAN I3 HAS ARRIVED! It was delayed twice and I got it!
F'prime
Veteran
I am also a C++ programmer (I am still learning it because i used C++ for Arduino and stuff)
SpeakCuba
Veteran
Arduino uses a mix of C# and C++ if I remember correctly, so Arduino C++ is slightly different and slightly easier than normal C++.
F'prime
Veteran
That is true.
FineGeoduck2211
Veteran
python is my main program lol, would you like to parter up or something?
edward_9x
Veteran
Here is how Cubing.js deals with a smart GAN cube:
cubing.js/src/cubing/bluetooth/smart-puzzle/gan.ts at main · cubing/cubing.js
🛠 A library for displaying and working with twisty puzzles. Also currently home to the code for Twizzle. - cubing/cubing.js
github.com
Justincubes
Veteran
It unfortunately involves reverse engineering everything, it seems like gan doesn't provide any documentation for this. Some parts would be pretty easy (like finding the hexadecimal that represents a move) but some parts are too complicated and I don't think you would be able to figure out why it works that way after studying the code for hours. You could probably just translate it line by line to your programming language of choice, or see if there is a cubing.js alternative for it. You could also switch to using node.js or deno for this project if you're familiar with JavaScript so you don't have to deal with the smart cube implemntation (but this would only work on a raspberry pi as arduinos don't support other programming languages afaik)
edward_9x
Veteran
Obviously, it goes without saying. Why would GAN make their technique public?
I wish there was a universal, formal standard for what smart cube packets represent. I also wish there was a similar standard for timer signals. Just a simple .pdf would save a lot of devs from having headaches.
F'prime
Veteran
Yo. Update time.
I just got my Raspberry Pi and i started the project. Unfortunately, my smart cube (GAN 356 i3) isnt showing up on Bluetooth on the OS. It is connecting to my phone when opening up CubeStation. I am using Raspberry Pi OS, which is basically a flavour of Arch Linux (i think? if im wrong please correct me). Has anyone ever had this issue before?
I just got my Raspberry Pi and i started the project. Unfortunately, my smart cube (GAN 356 i3) isnt showing up on Bluetooth on the OS. It is connecting to my phone when opening up CubeStation. I am using Raspberry Pi OS, which is basically a flavour of Arch Linux (i think? if im wrong please correct me). Has anyone ever had this issue before?
F'prime
Veteran
This is gonna be a long update.
Its been a while since i've said anything about the project, so here's a little update! I have managed to integrate a LLM (Large Language Model, basically what powers ChatGPT, Co-pilot and others) into a Raspberry Pi and managed to send the prompt to an Arduino Uno.
When I started the project, I first started with the Bluetooth part, but I ran into a problem. My smart cube wasnt connecting to my Raspberry Pi, no matter what I did. So, i skipped the bluetooth part and decided to go for the AI. Now, this part of the project, at least for me, was the hardest part of the project. I have never seen anyone actually do it, but ive managed to make it work. Here's how.
When I started working on the AI coach, I tried using ChatGPT's API. Most people usually use ChatGPT for these projects. But unfortunately, due to corporate greed, you had to pay some money to make it work, so I needed an alternative. Thankfully, ive found some alternatives like Google Gemini's API (yes you can use it in a Python project), but I didnt use it. Most of the API's either required money to operate, or werent supported by the Raspberry Pi 5.
But ive found a workaround. I used Ollama for this project. If you havent heard already, Ollama is basically a library for LLM's, meaning I am not limited to using just one LLM by one company. You can run it on Windows, Linux, or Mac. You can find LLMs from Microsoft, Google, Meta, even Ali Baba, which meant I had a variety of LLM's to choose from. After a lot of consideration, I chose Phi-3 by Microsoft. It was probably the most efficient out of the bunch, with a fast typing speed and efficient responses. But then I ran into a problem.
Most websites that used the Raspberry Pi to run Ollama used the Linux terminal to run their LLM's. After some digging, i found out that Ollama had an official Python library for integrating Ollama into Python. This was great news. I made a program where you had to insert a question, and the AI would answer it. All I needed to do now was add some code to the program that would send the AI's answer via serial communication to the Arduino to be shown on the OLED screen module (the small ones).
I've ran into a number of errors in my code because the prompt wasnt being delivered in string for some reason, then i rewrote the code. It worked from there. The OLED screen got short-circuited (i dont even know how that happened), so I wasnt able to use it, but after some tests, I found out that the Raspberry Pi was indeed sending what the AI was saying to the Arduino. It worked.
But here's why i'm writing this update. I am not moving forward with this project, at least for now. I am about to enter High School in the following weeks so I wont have time to work on the project. But, thats not saying it cant be done. Someday I might be able to make the project into a reality someday, but time will tell.
Its been a while since i've said anything about the project, so here's a little update! I have managed to integrate a LLM (Large Language Model, basically what powers ChatGPT, Co-pilot and others) into a Raspberry Pi and managed to send the prompt to an Arduino Uno.
When I started the project, I first started with the Bluetooth part, but I ran into a problem. My smart cube wasnt connecting to my Raspberry Pi, no matter what I did. So, i skipped the bluetooth part and decided to go for the AI. Now, this part of the project, at least for me, was the hardest part of the project. I have never seen anyone actually do it, but ive managed to make it work. Here's how.
When I started working on the AI coach, I tried using ChatGPT's API. Most people usually use ChatGPT for these projects. But unfortunately, due to corporate greed, you had to pay some money to make it work, so I needed an alternative. Thankfully, ive found some alternatives like Google Gemini's API (yes you can use it in a Python project), but I didnt use it. Most of the API's either required money to operate, or werent supported by the Raspberry Pi 5.
But ive found a workaround. I used Ollama for this project. If you havent heard already, Ollama is basically a library for LLM's, meaning I am not limited to using just one LLM by one company. You can run it on Windows, Linux, or Mac. You can find LLMs from Microsoft, Google, Meta, even Ali Baba, which meant I had a variety of LLM's to choose from. After a lot of consideration, I chose Phi-3 by Microsoft. It was probably the most efficient out of the bunch, with a fast typing speed and efficient responses. But then I ran into a problem.
Most websites that used the Raspberry Pi to run Ollama used the Linux terminal to run their LLM's. After some digging, i found out that Ollama had an official Python library for integrating Ollama into Python. This was great news. I made a program where you had to insert a question, and the AI would answer it. All I needed to do now was add some code to the program that would send the AI's answer via serial communication to the Arduino to be shown on the OLED screen module (the small ones).
I've ran into a number of errors in my code because the prompt wasnt being delivered in string for some reason, then i rewrote the code. It worked from there. The OLED screen got short-circuited (i dont even know how that happened), so I wasnt able to use it, but after some tests, I found out that the Raspberry Pi was indeed sending what the AI was saying to the Arduino. It worked.
But here's why i'm writing this update. I am not moving forward with this project, at least for now. I am about to enter High School in the following weeks so I wont have time to work on the project. But, thats not saying it cant be done. Someday I might be able to make the project into a reality someday, but time will tell.
Similar threads
