Mandala Messaging

Inspiration

Everyone has experienced the overwhelm of messages — unread notifications piling up, the anxiety of forgetting a reply, or the guilt of missing an important moment. At Hack Princeton, we wanted to address texting anxiety directly. We realized that while countless apps track productivity or screen time, few actually support mental health in communication. Mandala Messaging was born to bring mindfulness and positivity into digital conversations.

What it does

Mandala Messaging is a local AI-powered app that integrates with iMessage on macOS to help users communicate more mindfully.
It wraps around your messages to allow sending and receiving, while analyzing conversations for their emotional tone, sentiment, and overall positivity.
The app provides:

AI-generated message suggestions to improve empathy and clarity
Positive reinforcement insights to encourage healthy communication habits
Agent-based actions that let users create reminders or calendar events based on messages
A dashboard overview of emotional and behavioral trends over time — all processed locally for maximum privacy.

How we built it

We built Mandala Messaging as a hybrid Next.js + Python FastAPI application:

Frontend: Next.js web app (hp-app) providing the user interface and iMessage integration.
Backend: FastAPI (dedalus_backend) exposing endpoints for AI analysis, powered by Dedalus Labs tool-calling and Gemini LLM for natural language understanding. -IMessage Integration: Photon's IMessage-kit for interacting with local messages.
Database: SQLite, for storing contacts, AI analysis, and trends over time.
Integration: The app reads iMessage data, analyzes it through the backend, and returns insights and suggestions in real time.

Challenges we ran into

Integrating with iMessage: Photon's Kit made it very easy to interact with messages, but unresolved issues led to development challenges.
Tool chaining with Dedalus Labs: Balancing model latency and response quality for real-time conversation analysis was challenging.
Sentiment accuracy: Ensuring emotional insights were empathetic and not judgmental required careful prompt-engineering
Frontend-backend synchronization: Coordinating live message updates with ongoing analysis results was complex, especially with asynchronous events.

Accomplishments that we're proud of

Built a local, privacy-preserving messaging assistant from scratch in under 48 hours.
Successfully integrated AI-driven conversation analysis with live message handling.
Designed a mindful UX that promotes positive communication rather than productivity pressure.
Created actionable insights that genuinely help users manage texting anxiety.
Help users be more productivity in imessages.

What we learned

How to connect AI models to real-world emotional contexts without being invasive or prescriptive.
Practical challenges of combining frontend UX design, natural language processing, and local system integration.
The importance of privacy-first mental health technology — keeping user data safe while still delivering valuable insights.
How to work as a team under tight hackathon constraints while staying grounded in our shared mission.

What's next for Mandala Messaging

Expanding integration beyond iMessage to platforms like WhatsApp, Discord, and Telegram.
Developing a mobile iOS companion app for real-time sentiment and mindfulness prompts.
Adding personalized communication coaching and journaling features.
Exploring partnerships with digital wellness platforms and therapists to make Mandala Messaging a tool for everyday mental health.