SynapseFlow

Inspiration

Honestly? I was just drowning in browser tabs and random notes on my desktop! I realized how much time I wasted just trying to find that one piece of info or remember what an article was even about. I thought, "There has to be a better way to organize all this brain dump for myself." And then, SynapseFlow popped into my head during a late-night brainstorm.

What it does

Okay, so imagine I've got this huge article, or just a bunch of messy meeting notes from work. I just paste it into SynapseFlow. My AI instantly zips through it, gives me a super quick summary, and pulls out all the important stuff – like names, places, companies, key ideas. Then it saves it all neatly for me. No more scrolling forever or trying to remember where I saw that one thing!

How I built it

This is where it got crazy cool. I basically told Bolt.new what I wanted, like "make me a React front-end, a Node backend, connect it to Supabase." And it just... started building! For the brains of the operation, I hooked into Google's Gemini 1.5 Flash API. It's surprisingly good at understanding what I throw at it and spitting out summaries and those neat little tags. So, a lot of prompting Bolt.new and then fine-tuning Gemini's output was my process.

Challenges I ran into

Oh man, so many! Getting the AI to consistently extract exactly what I wanted, especially the structured data for tags, was a fun dance – lots of tweaking prompts. And sometimes Bolt.new's initial code needed a little nudge from me to fit my exact vision for the UI, or to properly handle the Supabase connections. Debugging those little backend hiccups when the AI didn't return exactly the JSON I expected was a classic hackathon struggle I faced!

Accomplishments that I'm proud of

I actually have a working app! Seriously, to go from scattered ideas to something that actually processes text, uses an AI, and saves it all in a database in this short time frame feels awesome. The fact that the summaries are genuinely useful to me, and those tags are mostly spot-on? I'm super proud of that AI integration. It's already a real time-saver for my own research.

What I learned

This sprint taught me the immense power of AI-assisted development tools like Bolt.new for rapid prototyping and deployment. It's like having a co-pilot that just builds code for me. I also gained significant experience in prompt engineering for complex AI tasks with Gemini, understanding how to coax specific, structured outputs from LLMs for my needs. Plus, the usual hackathon lesson: sleep is overrated, but learning by doing is everything!

What's next for SynapseFlow

So much for my personal use! First, I want to make it smart enough to see connections between all my saved documents – imagine seeing "this concept connects to that old note!" Then, I want to let myself pull in content directly from web pages or PDFs, not just copy-paste. I'm also dreaming of building out a really robust tagging and filtering system just for my own organization. The ultimate goal is for it to be my ultimate external brain.

Built With

• bolt.new
• gemini
• json
• react
• supabase
• typescript