Mindmaply

Mindmaply snapshot

Inspiration

Since we've all use LLMs previously and sometimes find it hard to navigate through the 1D output they produce, our team wanted a way to turn scattered notes into a clear, visual understanding. We pinpointed mindmaps as they feel natural for learning. Mindmaps have the advantage of being able to see a top-level overview of what we're doing. Plus, pairing them with AI makes it easier to explore and connect ideas quickly.

What it does

Mindmaply lets you create mindmaps, chat with AI about a topic, and see the map evolve as you learn. You can add, switch, and delete mindmaps, then use the chat to deepen or refine the structure. This in turn creates a visually coherent way for similar ideas to be connected even when copilot goes on a tangent from the original prompt subject.

How we built it

We used FastAPI framework, Python for the backend, a lightweight frontend with HTML/CSS/JS, and D3 for rendering mindmap graphs. We used Google Gemini mainly due to the free credits that were offered, but the project easily can be refactored to use other LLM APIs, either for cost effectiveness or depth of knowledge. The app keeps the mindmap list, graph view, and chat in sync so every interaction updates the current map and conversation.

Challenges we ran into

Synchronising across the three panels was tricky, especially when the user switches mindmaps or deletes blocks. We also had to manage async UI updates so the app stays responsive while waiting for AI responses. The Frontend and backend integration proved difficult as well when trying to make sure that actions that happened on the frontend linked to the correct backend functionality. Trying to get the LLM to produce a consistent output proved a challenge as well, but we managed to overcome all of these challenges.

Accomplishments that we're proud of

We built a clean, interactive UI that feels cohesive and fast, and we successfully connected AI chat with live graph updates. The result feels like a learning workspace rather than a simple chatbot. We took inspiration from Genio's style as we thought this would demonstrate the ability of this software to integrate into one of their existing solutions.

What we learned

We learned how to design around async workflows, keep UI state consistent, and build a system where visual structure and conversation inform each other. We learnt a lot on APIs and using APIs to help prompt LLMs.

What's next for Mindmaply

We want to add richer editing tools, better collaboration, and export options. Currently it's possibly to collaborate by sharing JSON files but a major improvement would be to use the cloud. Giving options to have multiple graphs in a single mindmap window could be good, but a further innovation on this would be to have dynamic links between different mindmaps. Another improvement would be to be able to link children laterally as well as vertically, improving the redability of very large diagrams. We also plan to improve AI prompting and use an MCP so the maps become even more structured and accurate over time.