This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
I built StudyBuddy AI a student-focused AI study assistant that runs on a local model.
I'm a Computer Technology diploma student, and my study routine is messy. Notes in one place, a search engine in another tab, an AI chat tool somewhere else, and random resources scattered in between. I wanted a single place where I can ask a question, get an explanation, and keep going without jumping around.
I built it for myself and my friends and classmates who face similar study challenges. I wanted to create something practical that could make studying a little easier, instead of building another project just for my portfolio.
What it does:
- Lets you chat with an AI assistant and ask study questions
- Gives explanations to help you understand a topic
- Offers a simple dashboard to access the study features
- Runs the AI locally, so it doesn't depend on a paid cloud API
It's not a huge platform. It's a practical tool made by a student, for students.
Demo
Code
Explore StudyBuddy AI Repository
The repository includes setup instructions for running the project locally. I also documented the Windows setup, since that's the environment I use to develop and test the project.
How I Built It
I started with a simple idea: a clean web app with an AI assistant behind it. The stack is intentionally simple.
Backend
- Python with FastAPI
- API endpoints that the frontend talks to
Frontend
- HTML
- CSS
- JavaScript
AI layer
- Ollama to run models locally
- An open-weight Gemma2:2b model configured for the project
- The backend sends requests to the local model and returns the responses to the UI
The flow is simple. The frontend sends a request to the FastAPI backend, the backend passes it to Ollama running the Gemma model on the same machine, and the answer comes back to the browser.
Challenges along the way
Building around local inference taught me things a hosted API hides from you:
- You have to think about your own hardware. The model runs on your machine, so memory and speed matter.
- Setup is a real part of the project. Getting Ollama, the model and the backend working together, especially on Windows, took effort, and that's why I wrote the setup docs.
- Responses won't always feel as fast or as polished as a big hosted service. That was a trade-off I accepted.
Why Does Open Innovation Matter?
For this project, open innovation made the whole idea possible.
- Experimenting with open-weight models: I could try a Gemma model directly instead of only calling a black-box endpoint.
- Running AI locally: The model runs on my own machine. Once it's set up, I don't need a paid API key just to try things.
- More control over the stack: I can change how the backend talks to the model, adjust the setup, and modify the project as I learn.
- Learning how inference works: Setting up local inference made me understand what's actually happening between a request and a response, instead of treating an API as magic.
- Less dependence on paid closed APIs: For a student project, not needing to pay per request is a big plus.
Local models aren't automatically better than hosted APIs. Hosted services can be faster or offer stronger capabilities for certain tasks, while local inference depends on your own hardware.
For me, building StudyBuddy AI with an open-weight model was a chance to learn by doing, understand local inference, and build something without depending on a paid API. That's what made this approach meaningful for my project.
Prize Categories
- Best Use of Gemma
Top comments (0)