@charliem

Note: even thoughit says it only took 1h i moved around folders and really took closer to 5h, look at screenshot.

What I made

I built the Math SAT Browser Extension Tutor (with cursor), a lightweight in-browser assistant that provides real-time AI-driven guidance on any SAT practice site. By running models entirely locally, the tool allows students to scan an infinite number of math problems without encountering subscription barriers or API rate limits. Also provides a cool plant that grows with you as you study to encourage staying focused. After study session khan academy links to topics scanned will be provided for further help.

What was challenging

LaTeX Corruption: Local models frequently produced "glued" prose or broken LaTeX syntax which required extensive regex-based repair modules to normalize text and equation formatting.

What we're proud of

User Engagement: The integration of session-based tracking and "plant health" focus mechanics provides a tangible and gamified incentive for students to stay on task.
Performance: By utilizing local Ollama models with a multi-stage extraction and cleanup process, we achieved a high-speed tutoring experience that functions locally without heavy reliance on external API latency.

How to test the project

To test the extension, you will need to clone the repository from the GitHub link.
Local Setup: Since the extension runs locally, ensure you have Ollama installed and the qwen2.5vl:7b and qwen2.5:7b models pulled. Then navigate to any SAT practice site of your choice, open the extension sidebar, and use the scan feature on a math problem to initiate the AI tutoring session.