Aura

fashionista who? you will be ;)

Comment

Inspiration

The vague desire to improve your fashion is hitting again 😨. Every time you see a cute fit on social media you think about it, but then when it comes time to put together an outfit you don't know where to start and each decision feels like a chore 😿😵‍💫😵.

Fashion content is one of the most consumed media categories — billions of saves, millions of engaged viewers, creators with massive reach — on the internet. But there's never been an AI layer that closes the loop between watching and wearing. You see an amazing fit on a creator whose vibe you love, and then what? Tab-switching, sizing guesswork, buried reviews, abandoned carts. Most impulses die in that friction 🙂‍↔️🙂‍↔️.

We wish we had the fashion sense of the Bratz dolls ✨. We wish we could magically transform like the members of the Winx Club 🧚. This weekend, we decided that we wanted to channel these Y2K icons, level up our style, and bring fashion to the masses.

Whether you have a "passion for fashion" 💃💅, have "nothing to wear" 🥀☠️, or simply want to offload the burden of choosing what to wear, Aura is here to put together an outfit that is so you while you chillax 😴.

We built Aura to close that loop. She transforms passive content — the Pinterest pin, the Instagram reel, the creator lookbook — into an active, personalized shopping action. The content becomes the input. The purchase becomes the output. The user just has to say yes. This is what the entertainment and fashion media pipeline looks like when AI is the bridge. ur aura: 📈📈📈

What it does

Aura is an AI shopping agent made in a React web app with Anime.js animations and Three.js 3D visuals — built to feel editorial, not transactional. Deployed on DigitalOcean

  • Scout agent — searches for fashion across trendy sites like Fashion Nova, Gap, Amazon, Musinsa, Lewkin and more via Firecrawl, continuously finding new products that match the user's vibe profile

  • Review intelligence agent — uses Firecrawl /interact to crawl real review pages, extract sizing comments and garment measurements, and build a fit verdict per product

  • Reasoning agent — K2 Think V2 takes all candidates, all review data, all sizing verdicts, and the user's full profile, and decides what's actually worth surfacing

When you're ready, Knot AgenticShopping completes the purchase without leaving the app!

How we built it

Works best in full screen (1920×1080) Backend

  • ElevenLabs — voice input and Aura's spoken persona (sounds like bestie)
  • Gemini multimodal — reads outfit images and social media posts into structured garment and style data
  • K2 Think V2 — core reasoning behind Aura Intelligence: sizing verdicts, candidate ranking, personalized recommendation scripts
  • Gemini Flash — fast structured extraction from product pages and review text
  • Gemini Nano — virtual try-on generation (user's photo + product photo composited)
  • Firecrawl — live product search, page scraping, and interactive review crawling. all asynchronous, each shopping link's analysis running in parallel
  • Knot TransactionLink — pulls past purchase history so Aura knows what you already own
  • Knot AgenticShopping — one-click autonomous checkout directly on the platform, within the user's set budget

Frontend: React, Anime.js, Three.js, CSS — with design assets from Canva Pro, Orchid, and Enter.pro

Challenges we ran into

  • We were planning on using the Daedalus VMs to run 24/7 agent swarm, but unfortunately it was down
  • Finding the right look for our project took some time, so it lead to a big rush towards the end of the hackathon
  • We struggled to implement a try on feature that display the outfit on the user but could not find a free solution that could run in a reasonable amount of time

Accomplishments that we're proud of

  • We're proud of the frontend design, which took trial and error to see what would look best. Since two of us are relatively new to hackathons, it was hard but fun to see our ideas come to life!
  • We are proud that we built something fun that we would actually use after this hackathon
  • We are proud that we followed through on our plan and completed almost all of the features that we wanted

What we learned

  • Just one more celsius!
  • Frontend is hard :( but i think we managed to build something aesthetic!
  • gemini free tier api limits are the devil. We spent 3 hours on trying to use vertex to get more gemini credits and use nano banana to try on but sadly our efforts were in vain

What's next for Aura

  • Implement try on feature.
  • Subscribe to a creator's vibe and Aura interprets their content library for your next haul!
  • Dropping a link (eg: Pinterest board) into Aura and receiving shoppable matches sized to body within seconds

Built With

Share this project:

Updates