SaveNet

See my write-up on Medium for more - https://medium.com/p/a862dfbeb226 Twitter post - https://x.com/hrmtsh2/status/2071730657194336641?s=20

Inspiration

I'd often save posts on Instagram, hoping to revisit them later. Such as eateries I would visit depending on my mood (whether I wanted Chinese, or Indian, near or far, etc.) Or I would save academic posts that I would want to refer later. Only to have to scroll through a messy heap of posts with no way to search through them in a way that captures their actual meaning.

What it does

SaveNet comprises a browser extension and full-stack web app that

How I built it

I had written major portions of the post scraping logic of the extension prior to the commencement of the hackathon. But I took this opportunity to build out an end-to-end user experience, replete with a web app with AI-enabled chat features (albeit still in a BYOK mode) using Vercel's v0 (which made a solid scaffolding of the Next.js project). I used AWS RDS (Aurora PostgreSQL) as it was easy to integrate and instantiate with Vercel and v0, and had inherent Titan Embed embedding for post embeddings, and pgvector store for the same.

Challenges we ran into

The key challenges were -

• determining what features to have on the extension-side vs. the web app. ended up going for a clean separation of concerns.
• picking the right feature set to focus on to prepare a most viable product as soon as possible
• bonus: editing the demo video took way too long (but was a fun experience) ## Accomplishments that we're proud of A product that feels like one I would use myself very regularly and that solves a problem that isn't being solved this well. Even though this is a novelty project, I feel like such projects can extend into unifying, customising, and enhancing our online interactions across websites. ## What we learned How to use v0, navigating AWS console as a whole (AWS RDS provisioning, Cognito-based auth in particular), Vercel deployment and management (although did have some prior experience with this) ## What's next for SaveNet SaveNet currently works for Instagram saves, X bookmarks, and Chrome browser bookmarks/reading lists. What to expect in the future —
• More types of content. Think of the vast amount of posts we’d browse and save across YouTube, Reddit, Wikipedia, etc.
• Option to get recommendations emailed periodically (such as for something new to do/somewhere new to go every weekend) and other exciting agentic capabilities.
• Ability to opt out of the AI-based features, or use a model of your choice with an API key, download library data (say, for use with local models), etc.
• Open to other ideas to make this project more useful! Please feel free to contribute code/issues to the project GitHub repositories (web app, browser extension), or to reach out with suggestions via email (or to just say hi!)

Built With

• aurora
• css
• html
• javascript
• nextjs
• postgresql
• react