An advanced, interactive AI Assistant platform featuring a premium WebGL-powered 3D wave background, client-side session persistence, and secure local API proxy capabilities.
- 3D WebGL Wave Simulation: Powered by Three.js with sophisticated post-processing pipelines, including
UnrealBloomPass, customFilmGrainShadernoise, andLuminosityHighPassShaderfor glowing aesthetics. - Fluid Keyframe Animations: Wave motion, transitions, and camera focal animations managed dynamically via GSAP.
- NVIDIA NIM AI Integration: Streamed AI responses powered by
deepseek-ai/deepseek-v4-flash(with automated fallback tometa/llama-3.1-8b-instruct). - Multi-Mode Reasoning:
- Standard Mode: General chat interactions with real-time text streaming.
- Planning Mode: Detailed structured planning for complex developer tasks.
- Deep Research: Comprehensive analysis and structured inquiries.
- Session Persistence & History: Local conversation caching via browser
localStoragedisplayed in either a responsive sliding sidebar or a central modal drawer. - UX Micro-animations: Interactive prompt typing suggestions, border glow gradients, and responsive navigation menus.
- CORS Proxy Backend: Built-in multi-threaded Python server that serves project assets and proxies requests to
integrate.api.nvidia.comto prevent browser CORS exceptions.
- Frontend:
- Semantic HTML5 & CSS3 Custom Properties (Glassmorphism & dark-theme design tokens)
- Vanilla JS / jQuery 3.6.0 (DOM control, event handling, AJAX orchestration)
- Three.js (WebGL renderer)
- GSAP (GreenSock Animation Platform)
- Backend:
- Python 3 (
http.server&socketserver.ThreadingMixInproxy wrapper)
- Python 3 (
- AI Engine:
- NVIDIA NIM Serverless APIs
├── .vscode/
│ └── launch.json # Debugger and compound configurations for VS Code
├── index.html # Main frontend entry point
├── style.css # Styling sheet (layout, variables, animations)
├── app.js # Frontend application driver and NIM API client
├── server.py # Local python proxy server
└── README.md # Documentation (this file)