Inspiration
We inspired in the fact that finding the perfect wallpaper to show to your friends, girlfriend or post in internet is a tedious, complicated and stressfull page jump of motivation loss
What it does
The page does scraping in about 6-8 diferent websites that keep and categorizes images and videos to use it like wallpapers, also we have a chat system that use and pre defined questionary to find your ideal wallpaper and if it doesn't works, it calls an IA system that reads the last conversation a guides the scraping with the correct keywords
How we built it
The springbot calls the main program in port 8080, that calls a class that starts the frontend, in the frontend you have 2 pages, home and browse, in home page you have a searcher that redirects you and your search to browse, browse takes that search and calls wallpaper service, this one searches the words in the h2 databes and if isn't found, they calls the web scrapper and adds the results to database, then it catches the database results and show it to the user
Challenges we ran into
We worked separatedly the front, IA and scrapper, so unite them was very dificult mainly front + back end, also the database integration was to complex and runing the page in a server was very dificult
Accomplishments that we're proud of
The page scraping works almost perfect, the main and browse page looks very beauty and the chat system is very cheap in terms of tokens
What we learned
the jsoup library was a big learning, also the endpoint system of springbot and the main learning is that and standar questionary can works much better than an LMM system in most part of cases
What's next for Persepolis AI
perfect the database, because at the moment is based in h2, that is inefficient and can ruin the memory of the system, also perfect and delete the non important results, correct the ip blocking of some webs, a better management of categorys and the monetization
Built With
- clouding
- css3
- html5
- java
- maven
- openrouteservice
- springboot
- tailwind
Log in or sign up for Devpost to join the conversation.