A "Second Brain" for your glasses. Remember names, recall conversations, and track your social life.
MindTrace is a tool to help you remember people and what you talked about. It uses a wearable HUD (like Meta Glass) to identify faces in real-time and a web dashboard to search through your past interactions.
Basically, it's for when you forget names or need to remember that one thing someone mentioned three weeks ago.
| HUD Interface | Dashboard |
|---|---|
 |
 |
| AI Summarizer | Contact Directory |
 |
 |
| Smart Reminders | Advanced SOS |
 |
 |
MindTrace is split into three parts:
A React 19 app to manage your contacts and see your stats.
- View contact history and relationship notes.
- Charts for social activity (built with Recharts).
- Interaction summaries.
- SOS/alert settings.
The interface for wearable devices or mobile browsers.
- Fast face recognition.
- Real-time speech-to-text (ASR).
- Simple overlays to identify people instantly.
A FastAPI server that handles the heavy lifting:
- InsightFace: For face analysis and matching.
- OpenAI Whisper: Transcribes ambient audio.
- RAG Engine: Semantic search using ChromaDB and Google Gemini.
- LangChain: Connects the LLMs to your memory data.
- 👁️ Face ID: Matches faces against your contact list.
- 🎙️ Background Tracking: Records and transcribes chats, then tags them to the right person.
- 🧠 Memory Search: Search your history with natural language (e.g., "What did John say about his new job?").
- 📅 Auto-Reminders: Creates tasks based on what people say (e.g., "Buy that book Sarah recommended").
- 🚨 SOS System: Quick emergency alerts with location sharing.
- 📊 Social Stats: Visualizes how often you're meeting people.
- Framework: React 19 + Vite
- Styling: Tailwind CSS 4
- Motion: Framer Motion + Lenis
- Maps: React Leaflet
- Icons: Lucide React
- API: FastAPI (Python 3.12)
- DB: SQLAlchemy + ChromaDB (Vector)
- AI Models: InsightFace, Faster Whisper, Google Gemini
- Logic: LangChain + ONNX Runtime
For the technical specifics on how we tuned the models and why we picked them, check out MODELS.md.
graph TD
A[Glass HUD / Client] -->|Video/Audio Stream| B[FastAPI Server]
C[Web Dashboard] -->|Data/Auth| B
subgraph AI Engine
B --> D[Face Recognition]
B --> E[ASR / Transcription]
B --> F[Summarizer / LLM]
F --> G[RAG / Vector DB]
end
subgraph Storage
B --> H[(SQL DB)]
G --> I[(ChromaDB)]
end
D -->|Match| A
E -->|Transcript| F
F -->|Insights| C
- Node.js 18+
- Python 3.10+
- uv (Python package manager)
- Google Gemini API Key
-
Clone the repo
git clone https://github.com/yourusername/mindtrace.git cd mindtrace -
Run the Backend
cd server uv sync cp .env.example .env # Add your GEMINI_API_KEY uv run main.py
-
Run the Dashboard
cd client npm install npm run dev -
Run the Glass HUD (Optional)
cd glass-client npm install npm run dev
MindTrace/
├── client/ # Web Dashboard (React)
├── glass-client/ # Wearable HUD
├── server/
│ ├── app/ # Routes & Logic
│ ├── ai_engine/ # FaceID, ASR, RAG
│ ├── data/ # Storage & DBs
│ └── main.py # App Entry
├── screenshots/ # Images for README
├── QUICKSTART.md # 10-min guide
├── API.md # Endpoint docs
└── CONTRIBUTING.md # How to contribute
MIT License. See LICENSE for details.
Built for people who need a little help remembering.