Inspiration

A Massively Open Online Moog. Using InfluxDB, Grafana Canvas, and the latest Generative AI tools, we have packaged a Docker Extension which recreates a Moog Instrument's Model15, modelling the UI with Svelte, and generating audio with either puredata or ffmpeg.

What it does

The main usecase (an interactive musical instrument) can be reconfigured to serve similar yet distinct usecases, E.g., building a shared virtual worlds, such as community gardens, etc. The goal is to create a general purpose tool for education. Your first quiz. Given what you know so far, why is the project named Mr. Frundles? Think step by step...

How we built it

The main components are a Docker Extension, Chrome Extension, Storybook, Chromatic. Under the hood. The main feature is the use of fine tuning of foundational models to build complex configurations from initially simple instructions. The application is controlled by a state machine which migrates configuration from a JSON representation of a restic repository, through docker config and docker secrets, to storybook args (using qs to format JSON as a querystring), and finally from there to SvelteFlow.

Challenges we ran into

There are many technologies floating around. Scope creep was an issue. Sometimes it felt like components were required. Other times it felt like choices made about which tools to use were conflicting with one another.

Accomplishments that we're proud of

The fact that very little code is required. One could build this project from scratch in a day using instructions and scaffolding provided by docker extension init etc. Also, this project could be reconfigured for many different use cases.

What we learned

How to use fine-tuning of foundational models for great success.

What's next for Mr. Frundles

Since the project can be reconfigured for many usecases, I will likely deploy this app in various marketplaces, including OpenAI's newly launched marketplace. Moreover, Canva's ecosystem, and Atlassian's Forge, to mention but a few, are all release candidates.

Built With

Share this project:

Updates