Idea Machine Backend

This code is meant to run a http server locally on a raspberry pi to orchestrate print requests coming from the idea-machine webapp (see here), handle bluetooth connections to the dices, administrate the labels of the dices, read the buzzer input and (pre-) generate ideas.

Prerequisites

These are the other necessary repositories:

https://github.com/technologiestiftung/idea-machine-frontend
https://github.com/technologiestiftung/idea-machine-dice

If you want to run this yourself, you also need a supabase project (to store ideas) and an openai api key (to generate ideas).

Install npm dependencies

If you are not in the project folder, cd into it and install dependencies:

cd idea-machine-backend && npm ci

Add the .env file

Create a .env file in the root of the project and add the needed variables:

cp .env.example .env

Run it

You can run the server with the following command:

npm run start

or the dev mode (watching the directory):

npm run dev

Architecture / Flow

The server uses multiple technologies to handle the different tasks:

Bluetooth to connect to the dices and receive side change events
GPIO to read the buzzer input
A REST API to:
- handle read/update dice labels,
- live generate ideas,
- pregenerate ideas,
- pick an idea from pregenerated ideas
- regenerate pdfs (in case of design changes)
- shutdown
- is alive check
WebSockets to update the connection status to the dice administration website, and listen for new printing jobs in the DB

The main flow can be visually represented as follows:

How are the ideas generated?

The ideas are generated using gpt-4o and dall-e 3. A prompt is sent to the gpt-4o model, and the output is sent to the dall-e model to generate an image. The idea text and idea image are then used to generate an HTML (postcard), which is then converted to a PDF (postcard).