Introduction to ACRYX
ACRYX is an open-core framework designed to build, deploy, and orchestrate Geo-Sentient AI Agents purely at the edge. It abstracts away the complexity of LBS (Location-Based Services) validation, federated learning, and spatial graph construction.
Quick Start
The easiest way to get started with ACRYX is using the official CLI tool. Ensure you have Node.js 18.0 or later installed.
acryx init my-spatial-app
cd my-spatial-app
acryx dev
Configuration
Agents are configured via the acryx.config.json file located in your project root.
{
"version": "2.0",
"edge_mode": true,
"agents": {
"sentinel": { "lbs_precision": "sub-meter" },
"predictive": { "llm_backend": "local-llama-3-8b" }
}
}Note on Edge Models
If your target device lacks sufficient memory for the 8B parameter model, ACRYX will automatically fallback to the quantized 3B model via federated sync.
Geo-Sentient Agents
Agents are the core operational units within the ACRYX ecosystem. Unlike traditional bots, Geo-Sentient Agents possess awareness of their physical hardware location, network topology, and immediate spatial surroundings (via LBS APIs).
Edge-Native LLMs
We ship heavily quantized, specialized Large Language Models (LLMs) ranging from 1B to 8B parameters. These models run via WebAssembly or native core ML runtimes directly on the edge node, ensuring 0ms network latency and zero data egress.
Spatial Graphs
A Spatial Graph is a real-time data structure representing the physical relationships between localized devices, POIs (Points of Interest), and network infrastructure. It is dynamically constructed and destroyed as devices move through physical space.
CLI Commands Reference
| Command | Description |
|---|---|
| acryx init <dir> | Scaffolds a new ACRYX project in the specified directory. |
| acryx deploy | Compiles and pushes the agent matrix to target edge nodes. |
| acryx sync | Force synchronizes the local LLM weights via federated nodes. |
Node.js SDK
For deep integrations, import the core library directly into your Node.js or edge runtime environments.
const matrix = new AgentMatrix();
await matrix.initialize();