reading_app

• 

  Application Usage

Inspiration

This project was inspired from tutoring a teenager who claimed to not have anything to read. They mentioned that libraries were archaic, and that they prefer simply getting everything from the internet, as such this concept was designed and constructed.

What it does

This application leverages the capacity of the OpenAI gpt-oss-20b models to dynamically serve as reading passage/article generation system. The 20b model was opted for, as it requires less resources to potentially run locally.

How we built it

This concept was initially prototyped using Bun, Typescript; some of the initial functionality was also constructed using Swift, but ultimately Egui Rust was opted for due to type-safety, wasm(web assembly) availability, and Cargo, Rust toolchain features.

Challenges we ran into

The concept encountered a number of challenges ranging from amount of time available to constructed instances, availability of Rust libraries intended for large language models, and distribution complexity.

Accomplishments that we're proud of

While the instance appears fairly trivial, I'm proud to have constructed a tool which could serve students and could be further improved upon by the open source community(and myself post judging)

What we learned

This project allowed me to better learn programming in Rust, had me learning Swift concurrency, and also had me studying the inner workings of large language models via mlc and web-llm.

What's next for cipher

After this endeavor, I will likely rest for about two days and then work on a fork of this instance involving abstract syntax tree traversal, incorporating Whisper for speech to text capacity, and potentially integrating ElevenLabs for a local agent users could converse with about the articles produced. I would also ideally like to improve the user interface and conduct data analysis on my own respective usage of the instance.

Built With

• egui
• groq
• rust