"Reasoning across decades, not days."
LifeGraph is a "Scientific Biographer" that solves Digital Amnesia. It captures the invisible 90% of your life (conversations, video, insights) and uses Gemini 3.0 to structure it into a searchable knowledge graph.
LifeGraph is built on the premise that LLMs are not just chatbots, but Semantic Kernels for databases.
- Voice-First Interface: Friction-free capture. Speak your thoughts, and Gemini structures them into
Events,Emotions, andDomains. - Video-to-Memory: Native multimodal ingestion. Stream raw video to Gemini 3.0 to capture non-verbal context (tone, environment) that text misses.
- Vector Memory (Pinecone): Every memory is embedded using
text-embedding-004and stored in a high-dimensional vector space. - Hyper-RAG: The "Ask" tab doesn't just keywords search. It retrieves semantic concepts (e.g., "Burnout patterns") across years of data and uses Gemini 3.0 to reason about cause-and-effect.
- Genesis Mode: An AI interviewer that helps you define your core values and 50-year vision, which then acts as the context for all future data analysis.
- Frontend: Next.js 14, TailwindCSS, Framer Motion (Universal Web App / PWA)
- Backend: FastAPI (Python)
- AI Models:
- Perception:
gemini-3.0-flash(Video/Audio/Image) - Reasoning:
gemini-3.0-flash(RAG Synthesis) - Embeddings:
text-embedding-004
- Perception:
- Database: Pinecone (Vector), JSON (Flat file storage for portability)
- Node.js 18+
- Python 3.10+
- Google Gemini API Key
- Pinecone API Key
-
Clone the Repo
git clone https://github.com/yourusername/lifegraph.git cd lifegraph -
Backend Setup
cd backend pip install -r requirements.txt # Create .env file echo "GEMINI_API_KEY=your_key" > .env echo "PINECONE_API_KEY=your_key" >> .env # Run Server python -m uvicorn main:app --reload
-
Frontend Setup
cd frontend npm install npm run dev -
Open App Visit
http://localhost:3000to start your LifeGraph.
- The Archive: Ingesting 20 years of Google Photos/Drive to build a retrospective graph.
- The Wearable: Serving as the backend for "Always-On" hardware pins.
Built for the Google Gemini Hackathon. Powered by Gemini 3.0.