Inspiration
Ever tried searching for a reel on Instagram, only to get results that have nothing to do with what you were looking for? Imagine being able to search for what’s actually inside a video, even when those words were never written in the caption. That’s the problem we wanted to solve with SaveYour.
What it does
SaveYour lets you save posts from Instagram, Reddit, TikTok, Twitter/X, and other platforms into one place. AI analyzes the content so you can search across all your saves semantically.
One cool example is when you’re going on vacation. You might have your list of go-to Michelin-starred restaurants from Reddit, while Instagram has all the tourist spots you want to visit. By having them all in one place, you spend less time switching between apps and more time planning your trip. It’s also collaborative, so your friends can add to the album too, instead of flooding the group chat with links.
How we built it
Our tech stack consists of:
- Flutter for our iOS and Android apps
- Snowflake Cortex as our AI pipeline. It uses the Gemini Live API to analyze the content and generate descriptions of the audio and images for semantic search
- Meilisearch for semantic search. When we get a description back from Snowflake, we store it so it can be searched semantically
- MongoDB as our main database
- Redis for caching database responses
- SeaweedFS for file storage
- Google for authentication
We also had an interesting infrastructure setup. Chakorn has a server running on his PC back home that we can access remotely, so we were able to run services like Redis on it and skip server costs entirely.
Challenges we ran into
The biggest challenge we ran into was setting up Snowflake, which took us 4 hours to configure from 1AM to 5AM. In the end, it was worth it. It fits neatly into our project and serves a huge purpose.
Accomplishments that we're proud of
We’re proud of the decision to not force ourselves into unnecessary prize tracks. Some of the tools simply didn’t apply to our project or fit our vision. Instead of forcing them into the stack just to qualify for another prize, we decided to focus on the technologies we absolutely needed. This saved us time that would have otherwise gone into learning and integrating technology that had no real application to us.
What we learned
We learned how important it is to adapt quickly during a hackathon. We started with a completely different tech stack, but once we realized it wasn't going to work for what we wanted to build, we immediately pivoted instead of trying to force it to work. That ended up saving us a lot of time and let us focus on actually building the product.
We also learned that being flexible with how you build something is important. The original plan doesn't always have to be the final plan, as long as you can recognize when something isn't working and make the change early.
What's next for SaveYour
One thing we want to add is Dynamic Classification.
The idea is to automatically add a new saved post to the album it best fits, without the user having to manually organize it. The AI would analyze the post, identify what it’s about, and compare it to the user's existing albums to determine where they would most likely have put it themselves.
By extension, we want it to be able to learn the names of new objects and concepts, making it better at understanding and organizing your saves over time.
Log in or sign up for Devpost to join the conversation.