Orma: Your AI-Powered Digital Memory
Orma is a smart memory layer for your browser that uses local AI to transform your scattered browsing history into a searchable, chat-ready personal knowledge base.
Inspiration
As a passionate learner who spends hours browsing the web, I kept running into the same frustrating cycle: I'd discover amazing information; an insightful article, a critical Stack Overflow answer, a fascinating research paper only to completely forget where I found it days later.
Browser bookmarks became a graveyard of forgotten tabs. Note-taking apps felt disconnected from the actual browsing experience. Nothing understood the context of what I was saving.
I wanted something different, a tool that could remember not just what I saved, but truly understand it, connect it to related ideas, and help me learn better over time.
That's when the idea for Orma was born: a browser extension that acts as your digital memory, intelligently understanding and connecting everything you learn online.
What It Does
Orma is a sophisticated memory layer for your browser that transforms how you capture and interact with online information. Powered by AI, it automatically processes and connects your browsing history, enabling you to:
- 💾 Save anything with one click: Capture pages, highlights, and snippets instantly without leaving your workflow.
- 🔍 Search using natural language: Ask questions like "that article about neural networks I read last week" and get results instantly.
- 💬 Chat with your browsing history: Have a conversation with everything you've ever saved, powered by contextual AI.
- 📝 Generate smart summaries: Get AI-generated summaries of saved content, distilling hours of reading into key takeaways.
- 📂 Auto-organize into project workspaces: Orma intelligently clusters related memories into workspaces without manual sorting.
- 🤖 Enhance AI tools with perfect context : Supercharge tools like ChatGPT by injecting relevant memories as context for richer, more accurate responses.
How We Built It
Orma is built on a carefully selected stack optimized for performance, privacy, and elegance:
| Layer | Technology | Why |
|---|---|---|
| Frontend | React 18 + TailwindCSS | Native-feeling, responsive interface |
| AI Processing | Google Nano AI | Lightning-fast local processing, no data leaves your device |
| Memory Storage | IndexedDB via Dexie.js | Reliable, high-performance vector storage in the browser |
| Search | Vector search (OpenAI Embeddings) | Semantic understanding, not just keyword matching |
| Extension | Chrome Manifest V3 | Modern, secure, and performant extension architecture |
The architecture is offline-first while maintaining real-time responsiveness. Every interaction provides immediate visual feedback with smooth animations, making the experience feel instantaneous.
Challenges We Ran Into
🗄️ Vector Storage in the Browser
Implementing efficient vector storage inside IndexedDB was uncharted territory. We had to carefully optimize storage schemas and retrieval mechanisms to maintain sub-second performance even as the memory database grew to thousands of entries.
⚡ Real-Time Processing
Balancing immediate user feedback with background AI processing was a constant tension. We solved this by implementing an intelligent queue system with granular visual feedback so the UI never freezes, and users always know what's happening.
🧠 Context Preservation
Ensuring saved memories retained their original context while remaining useful for AI interactions required multiple iterations of the embedding pipeline. A snippet means nothing without the page it came from we had to encode that relationship.
🪶 Extension Performance
Keeping the extension lightweight while handling complex vector operations and AI inference was tough. We leveraged Chrome's built-in AI capabilities (Nano AI) to offload processing and keep the memory footprint minimal.
Accomplishments We're Proud Of
- ✅ Built a fully functional memory system that genuinely enhances the learning workflow
- ✅ Achieved near-instant semantic search across thousands of saved memories
- ✅ Created a natural chat interface that lets you converse with your browsing history
- ✅ Implemented smooth, real-time UI updates with polished animations
- ✅ Designed an intuitive interface that stays out of your way until you need it
What We Learned
- Deep practical insights into vector databases and embeddings in constrained environments
- Techniques for high-performance browser extension development under Manifest V3
- Strategies for hybrid local + cloud AI processing pipelines
- The critical importance of user feedback loops in memory management UX
- Methods for preserving and leveraging context in AI-powered applications
What's Next for Orma The Ultimate Browser Companion
| Feature | Description |
|---|---|
| 🤝 Collaborative Memory Spaces | Enable teams to build shared knowledge bases imagine a team that never forgets a research finding |
| 📊 Advanced Learning Analytics | Surface insights into your learning patterns what topics you explore, knowledge gaps, and growth over time |
| 🌐 Cross-Browser Support | Expand beyond Chrome to Firefox, Safari, Edge, and Arc |
| 🔒 Enhanced Offline AI | Strengthen local processing capabilities for complete privacy and offline functionality |
| 🔌 API Ecosystem | Open up Orma's memory layer for developers to build on integrations, plugins, and custom workflows |
Orma is more than a project. It's a step toward enhancing human memory with AI in a way that feels natural and empowering. We believe technology should amplify our natural abilities, not replace them. Orma is our contribution to that vision.
Log in or sign up for Devpost to join the conversation.