yashindane/webcube-dualllm repository overview

⁠webcube

Rubik's cube assistant on Flask webapp. This webapp accepts the six faces of your cube and gives you the voice instructions as a response.

Demo -> Link⁠

Base Image -> Link⁠

⁠Usage

Navigate to http://<IP>:<PORT>/input. This webapp runs on port no. 85 by default, but can be changed in the app.py file. To use take edge to edge and centred pics of the cube. Start with Red face with the White face down, and take pictures in the order Yellow -> Green -> Red -> White -> Blue -> Orange. After this click get solution. While listening to the instructions face the Red centred face with the White centred face down.

⁠Running the container

sudo docker run -dit -p 5000:85 --name <container-name> yashindane/webcube-dualllm:v1 --apikey="<OPENAI/GEMINI API-KEY>" --mod="<openai/gemini>"

⁠Working

The six images of six faces have there respective data_uri, which are submitted by a form when you click get solution. This data_uri are converted to images and saved. Each image of the face of cube is processed by LLM, due to get the 9 tile colors. This tile colors build up the cubestring, which is then processed by the kociemba library to get the optimal solution to the cube.

⁠Cube String notation

colour_mappings = {
    "red": "F",
    "green": "R",
    "blue": "L",
    "yellow": "U",
    "white": "D",
    "orange": "B"
 }

The cubestring is passed to the kociemba.solve() function, which return a string containing instructions for solving the cube. Kociemba is a Python/C implementation of Herbert Kociemba's Two-Phase algorithm for solving Rubik's Cube.

Read full documentation of Kociemba here -> Link⁠

the instructions are decoded to human voice instructions. This instructions are then written to the output.js file, which outputs this instructions as voice.

⁠Building docker image

build docker image by ->

$ docker build -t <username>/<repo-name>:<version>

I have also uploded already build image for this webapp on Docker Hub -> Link⁠

⁠deploying the image in Kubernetes

create a deployment by ->

$ kubectl create deployment <deploy-name> --image <username>/<repo-name>:<version>

Scale the deployment if necessary and create a service by ->

$ kubectl expose deployment <deploy-name> --port=85 --name=<service-name> --type=LoadBalancer

⁠Prerequisite browser settings

⁠1. Remove chrome managed by your organisation (Optional)

Visit this link -> link⁠

⁠2. Edit chrome flags for camera access

Navigate to chrome://flags/#unsafely-treat-insecure-origin-as-secure on chrome browser and add http://<IP>:<PORT> inside Insecure origins treated as secure box and click enable and reload the chrome page. After usage just remove the entry and click disable.