Your campus. Your GPU. Your AI.
An offline-first, AMD-accelerated AI collaboration workspace for college students and campus teams. No cloud. No subscriptions. No internet required.
RyFlow turns the AMD chip already inside your laptop into a private, offline AI brain that your entire campus team shares. Every feature (writing assistance, semantic search, real-time collaboration, voice transcription) runs locally on your device using AMD's open-source ROCm stack and free, quantized language models.
The result: A zero-cost AI workspace that works when the WiFi goes down, remembers everything your team has ever created, and gets faster when it detects an AMD GPU.
See ARCHITECTURE.md for the full system design.
| Feature | Description |
|---|---|
| Local AI | Phi-3 Mini / Gemma 2B via Ollama - runs entirely on-device |
| AMD ROCm | Auto-detects AMD GPU and switches to hardware-accelerated inference |
| P2P Collaboration | Real-time co-editing over LAN via WebRTC + Y.js CRDTs |
| Knowledge Graph | Semantic vector search over all workspace content |
| Document Editor | TipTap rich text editor with AI writing assistance |
| Code Editor | Monaco editor with AI explain, debug, and optimize |
| Canvas | Excalidraw-powered freeform drawing with AI description |
| Task Board | Natural language task creation - type a sentence, get a Kanban card |
| AI Studio | Persistent chat with RAG - answers informed by your workspace |
| Study Guide | Auto-generate flashcards, key terms, and quizzes from your docs |
| Voice Input | Offline transcription via Whisper.cpp (Ryzen AI NPU accelerated) |
| Tags | Tag any item across the workspace for filtered views |
| Daily Notes | Auto-created daily scratchpad |
| Workspace Export | Export entire workspace as .ryflow file, import on any device |
| LAN Discovery | Auto-discover teammates on the same network via mDNS |
| Category | Technologies |
|---|---|
| Frontend | React 18 + Vite, TailwindCSS, TipTap, D3.js, Monaco Editor, Excalidraw, Framer Motion, Zustand |
| Backend | Node.js + Express, Socket.io, better-sqlite3 (SQLite) |
| AI Layer | Ollama (Phi-3 Mini, nomic-embed-text), Whisper.cpp, Pollinations.ai |
| AMD Stack | ROCm (GPU inference), llama.cpp HIP, Ryzen AI SDK + ONNX Runtime (NPU) |
| Networking | WebRTC (simple-peer), Y.js CRDTs, Bonjour/mDNS |
| Desktop | Electron |
| Container | Docker + Docker Compose |
The easiest way to run RyFlow. Downloads Ollama and the required AI models automatically.
Prerequisites: Docker Desktop
git clone https://github.com/sh4shv4t/RyFlow.git
cd RyFlow
docker compose upThen open: http://localhost:5173
First run downloads phi3:mini (~2.4GB) and nomic-embed-text (~270MB). This is a one-time setup. Subsequent starts are instant.
Prerequisites:
- Node.js 18+
- Ollama installed and running
# 1. Clone the repo
git clone https://github.com/sh4shv4t/RyFlow.git
cd RyFlow
# 2. Install dependencies
npm install
cd backend && npm install && cd ..
cd frontend && npm install && cd ..
# 3. Pull AI models (one-time)
ollama pull phi3:mini
ollama pull nomic-embed-text
# 4. Start Ollama
ollama serve
# 5. Run in development mode
npm run dev# After completing Option 2 setup:
npm run electronIf you have an AMD GPU, install ROCm for up to 10x faster AI inference:
- Linux: ROCm Installation Guide
- Windows: ROCm support via WSL2
Ollama will automatically detect and use ROCm.
RyFlow shows ⚡ AMD in the top bar when active.
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
make
bash ./models/download-ggml-model.sh base# Run backend and frontend together
npm run dev
# Backend only (port 3001)
npm run dev:backend
# Frontend only (port 5173)
npm run dev:frontend
# Run tests
npm --prefix frontend test -- --run
# Build for production
npm --prefix frontend run buildRyFlow runs Phi-3 Mini via Ollama entirely on your device. No data leaves your machine. Every chat, summarization, and task parse happens locally. When an AMD GPU with ROCm is detected, inference is automatically offloaded to the GPU.
Every document, task, code file, canvas, and chat is converted into a 768-dimension vector embedding using nomic-embed-text. These are stored in SQLite as binary Float32 buffers. Semantic search uses cosine similarity to find related content across your entire workspace history.
Teammates on the same WiFi network are discovered automatically via mDNS. Documents sync in real-time using WebRTC data channels and Y.js CRDTs - the same conflict-free merging algorithm used by Google Docs, but with no Google server in the middle.
Export your entire workspace as a .ryflow file (a zip containing
the SQLite database + uploads). Import it on any device to continue
where you left off. The institutional memory never dies when a batch
graduates.
Contributions are welcome. Please open an issue before submitting a pull request for large changes.
# Fork, clone, then:
git checkout -b feature/your-feature-name
# Make changes
npm --prefix frontend test -- --run
git commit -m "feat: your feature description"
git push origin feature/your-feature-name
# Open a Pull RequestMIT - see LICENSE for details.