ai_nn_controller Documentation

Welcome to the documentation for ai_nn_controller — an open-source, AI-native network controller framework for building intelligent control applications across heterogeneous network domains.

Overview

ai_nn_controller is a domain-agnostic framework for building AI-powered control applications that manage network equipment across any domain — optical, wireless, RAN, core network, and more. Key features:

Declarative App Definition: Define control apps with simple Python decorators
Automatic API Generation: FastAPI REST endpoints auto-generated for each app
MCP Tool Generation: AI agents (like Claude) can control apps via auto-generated MCP tools
Plugin System: Reusable, installable plugins give control apps typed access to external services (storage, model registries, monitoring)
Distributed Architecture: Nodes, register, and message broker communicate via ZeroMQ
Dynamic Discovery: Nodes and apps only need to know the register address

Quick Example

from ai_nn_controller.decorators.aic_app import aic_app
from ai_nn_controller.AicApp import AicApp
from ai_nn_controller.AicController import AicController

@aic_app(name="MyControlApp")
class MyControlApp(AicApp):
    aic_app_id = 1
    control_loop_update_time = 2
    read_measurements = {3: ["gain", "power"]}
    control_functions = {3: ["SET_GAIN"]}
    # cell_ids auto-generated: [3]

    @classmethod
    def process(cls, measurements):
        gain = measurements.get(3, [{}])[-1].get("gain", 0)
        if gain > 25:
            cls.add_command(("SET_GAIN", {"node_id": 3, "value": {"target_gain": 20}}))

if __name__ == "__main__":
    AicController(with_api=True).run()

User Guide

User Guide

API Reference

API Reference

Examples

Examples

Contributing

Contributing