Buzz

• 

  Buzz

Inspiration

The inspiration for Buzz came from a two simple observations:

1. We consume massive amounts of content online every day, but how much do we actually retain? Whether it's reading technical documentation, news articles, or educational content, I noticed that passive consumption rarely leads to deep learning.
2. We don't exactly have a proper digitalized question bank in Nigeria as a whole, schools (public & private), bootcamps etc and with the issue that came up with jamb recently, i thought a system like this would help move forward in the right direction.

What it does

In a nut shell, this product is an extension than generates a short quizz (for now, would add a dropdown for users to select the number of questions) based on the content of your webpage (the current page) then saves it in a database.

How we built it

It's quite easy, I created with a monorepo structure using Turbo to manage both frontend and backend development, then i:

1. Built the Chrome extension with proper manifest configuration and content script injection.
2. Implemented webpage content scraping using Chrome's scripting API.
3. Initially used OpenAI, then migrated to Google Gemini for this hackathon but as time pass, i want to add support to multiple LLMs, what this does for now is to generate the quizzes.
4. Added authentication with Supabase for user management and progress tracking.
5. Developed the interactive quiz interface with timing, scoring, and navigation.
6. Implemented quiz history and leaderboard functionality.

Challenges we ran into

1. Cost OpenAI API costs were escalating during development, json consumes a lot of token per request. Solution: Migrated to Google Gemini API, requiring complete response parsing restructure
2. Browser Security Restrictions Chrome extensions have strict content security policies Solution: Implemented proper content script injection and message passing between extension components
3. AI Response Inconsistency AI sometimes returned malformed JSON or unexpected formats Solution: Built robust JSON extraction with regex parsing and error handling

Accomplishments that we're proud of

It works lol, i guess that counts

What we learned

1. Database Design, this means a lot to me because i've been thinking on a product-level scale so that opened my eyes and mind to a lot of business decisions that goes on
2. Mastered working with multiple AI APIs (OpenAI → Gemini) and handling different response structures
3. Learned the intricacies of Chrome extension development, including content scripts, background workers, and popup interactions
4. Gained experience orchestrating a Deno backend, SvelteKit frontend, and browser extension
5. Learned to handle complex state across different components (extension popup, background scripts, and web APIs)

What's next for Buzz

Has to be more ai support as a fallback on production so there will be options when i run out of tokens then i'll make a more robust error catching system before i attempt to deploy to the chrome store. Then create the mobile version of it (learnt react native for this lol), then maybe reach out to small scale bootcamps to make use of the system, might make it open source, i haven't decided yet lol but yeah, this has been more of a learning experience for me

Built With

• css
• deno
• hono
• supabase
• sveltekit