Studyforge：本地生成測驗草稿並匯出 PDF 的評估工具

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

What I Built

I built Studyforge, a locally hosted assessment builder for teachers, professors, tutors, and other educators.

An educator enters course material or learning objectives, adds details such as the course, grade level, and student instructions, then chooses how many multiple-choice, short-answer, and long-answer questions to generate. Studyforge creates an assessment draft with an answer key, shows it in a browser preview, and exports a print-ready PDF. The PDF can be made as a student copy without answers or a teacher copy with the answer key on a separate page.

The intended users are educators who want to prepare classroom tests more quickly. Schools, tutoring centers, and individual teachers could use a tool like this, though Studyforge does not currently include billing or paid plans.

Demo

There isn’t a public deployment or video demo yet. You can run it locally by following the setup instructions in the README.

After setup, open http://localhost:8000 to try the assessment builder.

Code

https://github.com/aLok-1105/-studyforge

How I Built It

Studyforge is a Node.js and Express application. Its browser UI is served by the Express server, which also provides endpoints for generating assessment drafts and downloading PDFs.

For AI, it uses Ollama to run the Gemma 3 270M model locally. The app sends the educator’s source material to the local Ollama API, which returns the questions and answer key. PDFKit generates the downloadable assessment PDFs.

The model and dependencies need to be downloaded during initial setup. After that, question generation and the app’s main workflow can run offline on the same computer: no hosted inference API is required. The interface also loads Google Fonts when an internet connection is available; offline, it falls back to system fonts.

Why Does Open Innovation Matter?

Open-weight models and local inference make it possible to build an AI-assisted teaching tool without depending on a closed, hosted inference API. Educators can run the model on their own machine, avoid sending course material to a third-party AI service by default, and use the app without per-request hosted-model fees.

It also makes the stack more adaptable: users can configure a different Ollama model to suit their hardware or needs. Local inference does have trade-offs—the model must be downloaded first, generation depends on available hardware, and smaller models may produce content that needs careful review. Studyforge treats generated assessments as drafts, not as a substitute for an educator’s judgment.

Prize Categories

No partner categories selected. Studyforge uses Ollama and Gemma; I have not confirmed that either corresponds to an eligible partner category for this challenge.