classnotes-ai：以本地運行的 Qwen3:4b 將課堂 PDF 轉成温習筆記

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

classnotes-ai: Turn Lecture PDFs Into Study Notes With Local AI

As a college student, I know the struggle of having huge PPTs thrown at your face (your google classroom face of course) and being expected to read through all of it in a few days (we all know real study begins only a few days before the exam).
Yes, we can generate notes using Claude, ChatGPT, etc., but their token limits can be a big problem when it comes to large PPTs/PDFs.
Thus, I made classnotes-ai, a simple webpage that runs Qwen3:4b locally on your computer and generates notes from PDFs for free. (Export your slides to PDF first.) It can be a little slow for longer lectures, but that cost is worth it compared to dealing with token limits and account switching with various LLMs, or reading the PPT/PDF (and god forbid reading the actual book) yourself.

It could also work as a free study-aid service for classmates, or, with a small hosted proxy, a paid tool for students who don't want AI subscriptions.

Demo

https://bespoke-crisp-a720c5.netlify.app/

Requires Ollama running locally with ollama pull qwen3:4b. To let the page talk to Ollama, start it with:
OLLAMA_ORIGINS="https://bespoke-crisp-a720c5.netlify.app" ollama serve

Code

https://github.com/DRmadebot/Projects/tree/main/classnotes-ai

How I Built It

I built classnotes-ai around local AI inference rather than a hosted model API.

The core workflow involves:

Extracting text from uploaded lecture PDFs.
Splitting long documents into smaller chunks.
Sending those chunks to a locally running model through Ollama.
Generating structured study notes from the extracted content.
Letting you download the notes as Markdown or print them to PDF.

The project uses Ollama for local model inference and Qwen3:4b as the model I'm experimenting with.

One of the challenges is handling long documents without overwhelming the available system memory or making the application painfully slow. Local inference makes resource management an important part of the engineering, not just an afterthought.

Why Does Open Innovation Matter?

Open innovation is what makes this project possible.

I can experiment with open-weight models and local inference without needing to pay for every API request or build an expensive AI backend.

It also gives users more control over their data. With a properly configured local setup, lecture material can be processed on the user's own machine instead of being sent to a third-party AI provider.

For a student building a project on a limited budget, this makes a real difference.

More importantly, it lets me focus on the actual problem: building a useful study tool around existing AI capabilities rather than reinventing the model itself.

Open models make experimentation accessible, while tools like Ollama make running those models much easier for developers.

Thanks for checking out my project! If you try it, I'd love to hear your feedback.