Event Brain – The Offline AI Copilot for Events & Hackathons

"The event assistant that still works when the Wi-Fi dies."

🎯 Problem

Events and hackathons almost ALWAYS start with:

• ❌ No Wi-Fi
• ❌ No 4G/5G signal
• ❌ No access to schedules, speaker info, maps, logistics
• ❌ Can't even text your teammate to find them
• ❌ Standing outside in the cold waiting for signal

✨ Solution: Event Brain

A mobile AI assistant that works 100% offline, using on-device AI via Cactus.

You download the event bundle once, then the app lets you:

✔️ Ask anything about the event offline:

• "What's happening after 3pm?"
• "Where is the submission link?"
• "Which sessions are about on-device AI?"
• "Who are the judges and what do they do?"

✔️ Built from a local event knowledge base:

• Schedule
• Speakers
• Locations
• Rules, judging criteria
• Wi-Fi / logistics
• Submission links

All stored fully on the device (JSON).

✔️ AI answers using only offline data

We filter relevant parts of the JSON then feed it to Qwen3-0.6 running through Cactus locally.

• No cloud
• No latency
• No data leaving the device

The perfect "edge AI" demo.

⭐ Memory Master Track (Bonus Feature)

Users can also save:

• ⭐ Starred sessions
• 📝 Notes about people they meet ("Met Roman — Cactus founder, likes edge-first demos")

Then ask:

• "What starred sessions do I have after lunch?"
• "Who did I meet that works on fintech?"

These personal notes are stored locally as a shared memory for the model.

This directly hits Track 1: Memory Master.

🚀 Features

Core Features

• 100% Offline AI Assistant - Powered by Cactus on-device AI (Qwen3-0.6)
• Event Knowledge Base - All event data stored locally in JSON
• Zero Latency - Instant responses, no network calls
• Complete Privacy - No data leaves your device

Memory Master Features

• Star Sessions - Mark favorite sessions to remember
• Session Notes - Add notes about specific sessions
• People Notes - Remember who you met and what they do
• Contextual Memory - AI uses your notes to answer questions

UI/UX

• Nothing Phone Aesthetic - Dark, minimal, clean design
• Three Tabs - Assistant, Schedule, People
• Smart Context Filtering - Only relevant data sent to AI
• Beautiful Date Formatting - Human-readable times

🛠️ Tech Stack

• React Native - Mobile app framework
• Cactus React Native SDK - On-device AI inference
• Qwen3-0.6 - Small, fast on-device language model
• AsyncStorage - Local data persistence
• Android - Primary platform (iOS support available)

📦 Installation

Prerequisites

• Node.js >= 20
• React Native development environment set up
• Android Studio (for Android builds)

Setup

1. Install dependencies:

   npm install

2. Start Metro bundler:

   npm start

3. Run on Android:

   npm run android

📱 Usage

First Launch

On first launch, the app will:

1. Download the Qwen3-0.6 model to your device (one-time, ~600MB)
2. Cache it locally for instant future use
3. Load your event data from the bundled JSON

Using the Assistant

1. Open the Assistant tab
2. Ask questions like:
   • "What's happening after 3pm?"
   • "Where is the submission link?"
   • "Who are the judges?"
   • "What starred sessions do I have?"
   • "Who did I meet that works on fintech?"

Managing Schedule

1. Open the Schedule tab
2. Star sessions by tapping the star icon
3. Add notes by tapping "+ Add Note" on any session
4. View your notes - they appear below each session

Remembering People

1. Open the People tab
2. Tap "+ Add Note About Someone"
3. Enter name and note (e.g., "Roman - Cactus founder, likes edge-first demos")
4. Ask the assistant about people you've met

🎨 Design Philosophy

Event Brain follows the Nothing phone aesthetic:

• Pure black background (#000000)
• Minimal borders (#1a1a1a, #2a2a2a)
• High contrast text (#ffffff on black)
• Rounded corners (16-24px radius)
• Clean typography (system fonts, proper weights)
• No unnecessary elements

🔒 Privacy & Offline

• 100% Offline - Works in airplane mode
• No Cloud - All processing on-device
• No Tracking - Zero analytics or telemetry
• Local Storage Only - All data stays on your device
• No Network Calls - Except initial model download

🎯 Why This App is Perfect for This Hackathon

✔️ Real pain the judges experienced today

• No Wi-Fi → useless schedules → chaos

✔️ A genuine on-device AI use case

• Not a chat wrapper
• Not a toy
• A real problem solved by edge inference

✔️ Zero-latency, offline, private

• Exactly what Cactus wants to show off

✔️ Lightweight and achievable in hackathon time

• Local JSON
• Simple UI (chat + schedule list)
• Cactus LM call
• Small context filter
• Optional notes/star feature

✔️ Nothing judges will love the design

• Clean, minimal, dark UI → fits Nothing phone aesthetic

✔️ Future potential beyond hackathon

• Events, universities, festivals, conferences
• Even low-connectivity regions

📝 Event Data Structure

Event data is stored in eventData.js as JSON:

{
  event_name: string,
  location: string,
  wifi: { name: string, password: string },
  submission: { form_url: string, deadline: string },
  sessions: [
    {
      id: string,
      title: string,
      description: string,
      start_time: ISO8601,
      end_time: ISO8601,
      location: string,
      tags: string[]
    }
  ],
  speakers: [
    {
      id: string,
      name: string,
      role: string,
      company: string,
      bio: string,
      tags: string[]
    }
  ]
}

🔧 Configuration

Changing Event Data

Edit eventData.js to update:

• Event name and location
• Wi-Fi credentials
• Schedule sessions
• Speaker information
• Submission links

Model Configuration

Edit cactusClient.js to change:

• Model name (currently qwen3-0.6)
• System prompt
• Context formatting

🐛 Troubleshooting

Model Download Fails

• Check internet connection (needed only for first download)
• Ensure device has enough storage (~1GB free)
• Try restarting the app

App Crashes on Launch

• Clear app data and reinstall
• Check React Native version compatibility
• Ensure Cactus SDK is properly installed

AI Responses Are Slow

• First inference may be slower (model loading)
• Subsequent queries should be instant
• Check device performance (older devices may be slower)

🚀 Building for Production

Android APK

cd android
./gradlew assembleRelease

The APK will be in android/app/build/outputs/apk/release/

iOS Build

cd ios
pod install
npm run ios

📄 License

This project is built for the Cactus x Nothing x Hugging Face Hackathon.

🙏 Acknowledgments

• Cactus - For the amazing on-device AI SDK
• Nothing - For the design inspiration
• Hugging Face - For the hackathon platform

---

Built with ❤️ for events that have no Wi-Fi