AURA (AI Universal Response Assistant) is a desktop AI assistant built with React, Tauri, Ollama, and Google Gemini Vision. It combines local LLM inference with cloud-based vision capabilities to provide a lightweight, privacy-focused desktop assistant through a floating overlay interface.
- 💬 Local AI conversations using Ollama
- 👁️ Screen understanding with Gemini Vision API
- 🎤 Voice input using Speech Recognition
- 🔊 Text-to-Speech responses
- 🖥️ Floating desktop assistant overlay
- 🧠 Conversation memory for contextual responses
- 📸 Screen capture and analysis
- ⚡ Hybrid AI routing (Local + Cloud)
- 🖱️ Draggable always-on-top assistant
- React
- Vite
- JavaScript
- Tauri
- Rust
- Ollama
- Gemma 3 (Local LLM)
- Google Gemini Vision API
- SpeechRecognition API
- SpeechSynthesis API
- Screen Capture API
src/
├── components/
├── services/
├── assets/
└── App.jsx
src-tauri/
├── src/
└── tauri.conf.json
git clone https://github.com/Yash-Rasal/AURA-Desktop-Assistant.gitcd AURA-Desktop-Assistantnpm installCreate a .env file.
VITE_GEMINI_API_KEY=YOUR_API_KEYDownload and install Ollama.
Pull the required model:
ollama pull gemma3:1bStart the Ollama server.
npm run tauri devScreenshots and demo GIF will be added soon.
- Multi-agent workflows
- Better memory management
- Plugin architecture
- OCR support
- Local vision model support
- Automation tools
- Cross-platform support
This project is intended for learning and portfolio purposes.