WiseTok

• 

  Sign in

• 

  Sign up

• 

  Main view - after uploading a file

• 

  Main view - after uploading a file (light mode)

• 

  LLM config

• 

  File uploading

• 

  File list

Inspiration

I've been rotting my brain with tiktok for quite a while. Then i realised: why i cannot doom scroll to actual knowledge instead of a new dance?

Then, WiseTok was born: a platform where you can doom scroll your knowledge, so instead of rotting your brain, you rotten your ignorance.

What it does

It uses AI & other tools to receive a PDF (or text file), process it, and extract some "bites" of knowledge from the text contents. You can pass a custom query to it, so you can extract conclusions from it. The limit is: the quality of the LLM selected (you can select & BYOK) and the quality of your content (the more OCR-able, the better)

How we built it

We used state of the art AI's to code parts of it, with close guidance from me, a Sr BE dev (with no frontend experience).

It combines a set of tools to process files, convert them to a format ingestable for agents and then a custom made prompt useful for extracting the most important parts of the content.

Then, those are presented on a Zen UI, with minimal distractions, so you can learn without thinking you are. The app has a social feature where you can share the extracted bites with other users, so everyone can learn. Twitter/X sharing is available too, as the format allows for "small wisdom" to be shared.

Challenges we ran into

AIs are bad at OCR. They don't like PDFs. So we had to resort to open source tooling that enabled us to process them before the AI processes it. This allowed the LLM to generate more "intelligent" content, with fine tuning from the user

Accomplishments that we're proud of

First "vibe coded" app. I put more focus on the safety of it, as users may upload private files to private llms (you can, too). Rate limiting, pressure metrics for ELU (node) and other cool backend stuff. Processing of files is "queued" as well, allowing UI clients for asynchronous processing of multiple files, just limited by the speed of the LLM they use

Best of all: ad free.

What's next for WiseTok

More social features, an actual deployment to the real world and faster models so wisdom is generated faster, better and stronger

Built With

• node.js
• ollama
• sqlite
• vite