📄 Technical Specifications & Architecture

Librae System Documentation

Geospatial intelligence infrastructure requirements, offline installation sequences, and zero-egress cybersecurity compliance protocols.

1. Hardware Specifications

CAHAYA executes high-fidelity point-cloud modeling and local LLM reasoning entirely on client hardware. The host machine must satisfy the following minimum and recommended requirements:

Recommended Host Configurations
⚡ CPU: Intel Xeon / AMD Threadripper (8+ cores) or Apple M-Series Max/Ultra.
⚡ RAM: 32 GB Minimum (64 GB or 128 GB recommended for multi-site simulations).
⚡ Storage: 250 GB NVMe SSD (reserved for point-clouds and local weights).
Local GPU Scaling Tiers
🚀 Cahaya Edge (7B model): 1x NVIDIA RTX 4060 (8GB VRAM) or Apple M-series (16GB Unified).
🚀 Cahaya Standard (32B model): 1x NVIDIA RTX 4090 / RTX A6000 (24GB+ VRAM).
🚀 Cahaya Enterprise (72B model): 2x NVIDIA H100 / A100 (144GB+ total VRAM).

2. Step-by-Step Offline Installation

For deployment in high-security, air-gapped environments, follow this installation sequence:

01
Extract Standalone Archive

Extract the compiled package to `/opt/lenuda/` on Unix-like operating systems or `C:\Program Files\Lenuda\` on Windows workstations.

02
Run Virtual Environment Initialization

Execute `chmod +x install.sh && ./install.sh` (Unix) or run `install.bat` as Administrator (Windows). This configures local virtual environments and seeds embedded local background API engines.

03
Seed Offline Model Weights

Download and place the designated model weights directly into the `models/` directory of the application root.

04
Execute Hardware Handshake

Bind the workstation license key by running: `python3 lenuda_cli.py --status`. Record the unique SHA-256 System Hardware Fingerprint and transmit it to Librae AI Labs to receive your annual activation token.

3. Cybersecurity & Compliance Enforcement

CAHAYA complies with strict government and defense security directives regarding data privacy and zero-trust computing:

Zero-Egress Isolation

The application binds locally to loopback interface `127.0.0.1`. It programmatically intercepts and blocks all outbound external network calls.

Immutable File Ledger

Files ingested into the simulation lab (like LiDAR points or raster layers) are validated against local SHA-256 Merkle root receipts. Any file modifications cause instant integrity failures.

4. High-Fidelity 3D Simulation Rendering

Immersive point-cloud and physics simulations are executed on the host's GPU using the Librae Core 3D Rendering Engine. By running a secure headless rendering system, CAHAYA streams a high-framerate rendering frame (30 FPS) directly to the local rendering container over secure local WebSockets. WGS 84 spatial coordinates are preserved natively via deep spatial mappings.