Ask-ED

Inspiration

We created Ask-ED to solve a common problem: students waste too much time searching through Ed Discussion forums for answers. We wanted to build an AI tool that makes finding relevant information quick and easy.

What it does

Ask-ED is an AI-powered chat interface connected to a database containing all of a user's Ed Discussions. It allows students to ask questions in natural language, automatically searches for relevant threads, provides synthesized answers, and displays direct links to the original discussions in a side panel. Syncing new messages is seamless and handled with a simple button. However, the user needs their Ed token, which can be difficult to obtain for casual users. That’s why we built Ask-ED Token Retrieval, a chrome extension that automatically fetches the token and places it where it needs to be.

How we built it

We built Ask-ED using Next.js, TypeScript, Plasmo, Neon for the database and AI technologies : text generation and embedding. We integrated with Ed Discussion's API to access course content to place in our database and developed a custom sidebar component to display source references. We used intercepted API calls to capture search results and display them as clickable links. For easy setup of the user ed token, a chrome extension is provided.

Challenges we ran into

The biggest challenge was reliably capturing and storing Ed Discussions : whether courses, threads, or answers, it was seriously tough. The Ed API is poorly documented and unreliable, so we had to rely heavily on reverse engineering. In the end, we found that the most effective solution was to use the user token and simulate browser behavior, rather than trying to work with the developer API. This approach proved to be much more stable and reliable.

Accomplishments that we're proud of

We’re especially proud of our automatic sync process, which pulls all the data into our local database. We also built a browser extension that keeps the experience smooth and user-friendly, and of course, our UI, which absolutely slaps.

What we learned

We learned effective techniques for intercepting API calls, parsing complex response formats, and designing fluid UI animations that enhance the user experience without being distracting. We also gained hands-on experience building a Chrome extension, setting up a local database and API, and integrating large language models (LLMs). To improve search relevance, we used embedding models to compute similarity scores between user queries and discussion content.

What's next for Ask-ED

Currently, the deployment of our app on https://ask-ed.ch isn’t fully functional yet, but the local version runs perfectly. Finalizing the deployment is one of the next steps we need to take. Looking ahead, future improvements include adding thread content previews in the sidebar, enhancing thread title extraction, and implementing a rating system to help users identify the most valuable discussions. We also plan to introduce course-specific learning optimizations. Although we didn’t have time to fully implement image support, all images have already been extracted from Ed, stored in our database, and are ready to be integrated. We believe this is the natural next step forward.

Built With

• bun
• neon
• next.js
• plasmo
• typescript