A modern web platform and foundation model suite pairing frontier AI deduction with traditional Vedic sacred geometry, animated GSAP flowing fluted glass heroes, and verifiable open weights.
Explore Models • Architecture • Quickstart • Docker Deployment
CSky AI provides transparent, auditable open weights designed for inspectability and sovereignty:
| Model | Sanskrit & Scale | Role & Architecture | Color Theme | License |
|---|---|---|---|---|
| Akash | आकाश · 405B | Flagship multi-step reasoning, formal theorem verification & deep logic | Celestial Rose (#E11D48) |
Open Research |
| Agni | अग्नि · 120B | High-throughput code synthesis, compiler verification & autonomous swarms | Solar Flame (#EA580C) |
Apache 2.0 |
| Drishti | दृष्टि · 70B | Enterprise production workhorse balancing latency, cost, and depth | Golden Amber (#D97706) |
Apache 2.0 |
| Prithvi | पृथ्वी · 32B | Grounded multimodal foundation for 3D spatial reasoning & 4K documents | Terracotta Bronze (#C2410C) |
Apache 2.0 |
| Soma | सोम · 14B | Distilled Lean 4 theorem proving & Olympiad-level mathematical logic | Lunar Indigo (#4F46E5) |
Apache 2.0 |
| Vayu | वायु · 8B | Ultra-compact edge model running offline on consumer laptops & devices | Mint Sage (#059669) |
Apache 2.0 |
- Canvas-driven GSAP Harmonics: Real sinusoidal wave generation customized per model (e.g. Akash Cosmic Ether Pillars, Drishti Prismatic Eye Focal Axis, Vayu Kinetic Fluttering Reeds).
- Vedic Sacred Geometry Watermarks: Intricately etched background mandalas and yantras featuring authentic geometric topologies (Surya Mandala, Akash Cosmos, Drishti Eye Yantra, Agni Panchagni Triangles, Prithvi Bhupura Fortress, Soma Chandra Crescent).
- Optical Depth-of-Field Refraction: Multi-layer composite with ambient halo glow, optical dispersion, and masked crisp focal geometry.
- Replaces static code dumps with an interactive 4-layer transformer topology:
- Rotary Position Embedding (RoPE) with 200k context sequence length.
- Grouped-Query Attention (GQA) with 8 KV heads for memory efficiency.
- SwiGLU Non-Linear Reasoning Engine with intermediate projection expansion.
- Native Block Quantization supporting FP8 and 4-bit inference kernels.
- Real-time interactive deduction visualizer showcasing token generation speed, cognitive latency, verified proof steps, and safety bounds.
- Dynamic navbar CTA with CSS transitions that harmonizes color palettes with active models (
#E11D48for Akash,#EA580Cfor Agni,#D97706for Drishti, etc.). - Comprehensive Mega-Menu dropdown with instant access to model specs and software tooling.
- Framework: Next.js 16 (App Router, Turbopack, React 19)
- Styling: Tailwind CSS v4
- Animation: GSAP (GreenSock Animation Platform)
- Typography: Geist & Geist Mono
- Deployment: Docker Multi-stage Standalone (
output: "standalone")
- Node.js 20+
- npm 10+ (or pnpm / yarn)
# Clone the repository
git clone https://github.com/CSky-Developments/AI.git
cd AI
# Install dependencies
npm install
# Start local development server
npm run devVisit http://localhost:3000 to view the application.
The application is pre-configured with output: "standalone" in next.config.ts for minimal production Docker container images:
# Build the Docker image
docker build -t csky-ai .
# Run the container
docker run -p 3000:3000 csky-aiThe server binds to 0.0.0.0:3000 with production optimization.
├── public/ # Static assets, branding & SVG emblems
├── src/
│ ├── app/ # Next.js App Router (layout, page, globals)
│ ├── components/ # Modular UI components
│ │ ├── views/ # Overview, Models, Manifesto, EarlyAccess, ProductView
│ │ ├── AnimatedBarsBackground.tsx # GSAP fluted glass wave canvas
│ │ ├── TraditionalSymbols.tsx # Sacred geometry SVGs & watermarks
│ │ ├── ModelArchitectureDiagram.tsx # Interactive 4-layer topology
│ │ ├── InteractiveReasoningDemo.tsx # Chain-of-thought deduction trace
│ │ ├── BenchmarkComparisonChart.tsx # Animated benchmark evaluation
│ │ ├── Navbar.tsx # Adaptive color-shifting navigation
│ │ └── Logo.tsx # Dynamic 4-bar brand logo
│ └── data/ # Product and model configuration metadata
├── Dockerfile # Multi-stage standalone container specification
├── next.config.ts # Standalone output & Turbopack configurations
└── package.json
Open-source under the Apache 2.0 License. Models released under their respective open weights licenses.