ARDI Documentation - agency

Agency Algorithms

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency Algorithms Algorithms represent the brain of the AI and are one of the key parts of the anatomy of an agent. There are a range of pre-made algorithms that allow you to create weak and strong AIs.

Agency Command-Line Parameters

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency Command-Line Parameters ParameterDefaultMeaning--serverlocalhostThe ARDI server to read data from--sitedefaultThe ARDI site to read data from--port9990The HTTP port for the JSON API--ua0Enable OPC-UA. See below for more--uaport4840The OPC-UA port to open

Agency Page

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency Page All of your active Agency alerts are available through the Agency web page. This can be accessed from the sidebar of your ARDI server. The Agency web-page updates every 10 seconds with the latest new Agency detections. There are two main lists of Agency alerts - the

Example Agents

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Example Agents Example - Comparing Actual to Target Example - Detecting Leaks or Discontinuities

Agent Walkthrough - Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agent Walkthrough - Attachment In the previous section, we set the rules for our weak AI so that it reports when a value exceeds 100%. We still haven't told it which assets to operate on yet though, or which properties to use. This is done in the attachment and binding steps, shown below.

Designing Your Agent

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Designing Your Agent Before we start building an agent, we need to design it - to decide what it will do, and what it will do it on. Let's start with a fairly basic example. In our database, we an asset called the Main Pump, which has the property

Agent Walkthrough - Settings

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agent Walkthrough - Settings In the previous step we decided what kind of agent we wanted, and set up the inputs and the algorithm. Now we need to fill in the settings for this agent, to make the actual decision behind the AI. Data/Settings The Data section gives the 'PropertyPattern' brain we've chosen some parameters. You can find out which parameters a pre-made algorithm supports by viewing its page on this wiki.

Alert Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Alert Attachment Alerts are atttached to assets in your ARDI database. There are several different methods of attachment... Direct / Asset Attachment In this case, your alert is attached directly to a single, named asset. Type Attachment Your alert will be attached to

Algorithms

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Algorithms An algorithm in Agency is the 'brain' of your agent. It defines the rules or the method that you're using to detect an issue. There are a range of built-in Agency algorithms, and you're also free to create your own using Python. Some of the pre-made algorithms are

The Anatomy of an Agent

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

The Anatomy of an Agent In AI, an agent is a single, self-contained and independent artificial intelligence (either weak or strong) that performs a task. In Agency, each alert you create is powered by an agent. Each agent has several key parts... Inputs

Agency APIs and Integration

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency APIs and Integration There are three primary ways to access your Agency detections. * Agency ARDI Addin * Agency JSON API * Agency OPC-UA Output

Agency ARDI Addin

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency ARDI Addin You can view Agency detections at any time by opening a browser and going to your ARDI server. You can find the 'Agency' menu on the left-hand sidebar. Click on it to open the Agency Detections page.

Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Attachment Attachment describes how your AI is attached to ARDI assets. The Algorithm, Inputs, Outputs and Data of your AI describes what the AI is, how it works and what it does. The Attachment and bindings of your AI discribe how it works with ARDI.. Attachment Types

BadDataDetector

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

BadDataDetector The baddatadetector algorithm returns a 1 when one or more input channels contain bad or null data, and otherwise returns 0. Usage This algorithm works with a minimum of one input. Parameters This algorithm has no properties or options.

Binding

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Binding Binding is the act of linking the inputs of your AI to the properties of the ARDI assets you've chosen to attach to. The binding phase always comes after attachment. A new copy of your AI is generated for each asset you want to connect to, and then they are bound to live data one-by-one.

Creating Your Own Agency Algorithm

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Creating Your Own Agency Algorithm It's quite possible to add your own algorithm with Agency. NOTE: Below is technical information for Python programmers Create a new file in the Agents folder of your Agency installation. Name it 'agent_your_algorithm_name

Differential

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Differential The differential algorithm calculates the greatest difference between all of its inputs and returns a 1 it exceeds a given min threshold or a 0 otherwise. Usage This algorithm is often used with two inputs, but can take as many as you like. It always compares the

dynamic

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

dynamic The root Agency web API endpoint used to make Agency available to ARDIs Dynamic Live Alert system. This allows ARDI to access Agency alerts and map them onto assets without needing the individual alerts configured in ARDI. Bridle #2 Roller 2 Exit Accumulator Prime Oven Zone 3 Not At Target PMT Finish Oven Zone 3 Not At Target PMT

Example - Comparing Actual to Target

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Example - Comparing Actual to Target This Agent has two inputs - the Actual value and the Target value - and uses the differential algorithm to flag if they have drifted apart from one-another. We want to raise an alert if they differ by more than 1.5 units.

Example - Detecting Leaks or Discontinuities

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Example - Detecting Leaks or Discontinuities This Agent has two inputs and is designed to spot a sudden drop in value across a sequence. In this case, we are looking for a difference between the value of a property between an asset and it's neighbour. We can do this with a

Extending the Example to Cover Multiple Assets

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Extending the Example to Cover Multiple Assets If you've followed the example so far, you might have felt that it's a very complex process for such a simple alert. And you'd be right. The complexity starts to become worth the trouble once you start using different

Agency Folders

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency Folders Your Agency installation folder contains a couple of folders.... Root Folder The root folder contains the actual Agency library, and often additional files such as configuration files and dependencies. Agents The agents folder contains your

Alert Full Names

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Alert Full Names The full name of the alert is only created after it's been attached to an asset. Since a single alert might be attached to dozens of different assets, the full agency name of an alert is a combination of both the name of the asset and the

grafana/all

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

grafana/all The root Agency web API endpoint used to make Agency to Grafana by delivering it as a Grafana-style table. This specific endpoint displays the status of all of your different alerts. [{"type":"table","columns":[{"text":"Name","type":"string"},{"text":"Status","type":"boolean"}],"rows":[[ "Very High Current Usage @ Entry Accumulator Status",0 ]}]

grafana

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

grafana The root Agency web API endpoint used to make Agency to Grafana by delivering it as a Grafana-style table. This specific endpoint displays the status of the active Agency alerts. [{"type":"table","columns":[{"text":"Name","type":"string"},{"text":"Status","type":"boolean"}],"rows":[[ "Very High Current Usage @ Entry Accumulator Status",0 ]}]

Input Names

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Input Names Agency inputs follow the format.... . For example, if you're looking for the Pressure (a property) of a cars Front Right Tire (an asset), you'd write... Front Right Tyre.Pressure Dynamic Names When you're attaching an alert to more than one asset, you'll need to dynamically update the name of the point you're attaching to.

Agency

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency Agency is the ARDI framework for deploying agents - small, flexible artificial intelligences designed to detect or predict issues.

JobLibAI

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

JobLibAI The JobLibAI algorithm loads a AI that has been saved using the JobLib Python module and runs it against the input data. Usage This algorithm is often used with AIs that have been created with Cognition, to provide live alerting of issues or predictions.

JSON API - List

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

JSON API - List Method: GETEndpoint: /listParameters: None Description Returns a list of all of the agents available in Agency. Response Returns a JSON-encoded payload. [{ "name": "Detection Type @ Detection Asset" } ] The response is an array of JSON objects that only include following fields

JSON API - Root

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

JSON API - Root Method: GETEndpoint: /Parameters: None Description Returns a list of all of the active Agency detections. Response Returns a JSON-encoded payload. [{ "name": "Detection Type @ Detection Asset", "element": "Output_Name", "priority": "2", "agent": "Detection Type", "triggered": "2021-04-30 12:12:04.032000" } ]

JSON API - Status

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

JSON API - Status Method: GETEndpoint: /status/Parameters: None Description Returns detailed status on a single agent. Parameters A url-encoded string should be included after /status/ in the URL. If an agent with that name exists, you'll be given extended details about it.

Agency JSON API

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency JSON API You can access the JSON API on port 9990 (unless you have changed the defaults on the command-line). Using a standard REST/HTTP GET request. There are three main endpoints... / (Root - Show All Active Alerts) /list (Show All Alerts) /status (Show Individual Alert Details)

list

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

list The list Agency endpoint provides a list of all of the alerts available in Agency, and their current status. [{ "name": "Out Of Temperature Control @ Finish Oven Zone 1", "status": "OK"}]

Normalising Inputs

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Normalising Inputs Agency - along with some other ARDI products - uses a Normalisation Engine to convert the inputs you're getting into inputs suitable for feeding an AI. In most cases, AIs need normalised inputs, meaning that they all have a similar numeric range.

OPC-UA Endpoint

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

OPC-UA Endpoint Agency includes an OPC-UA server to provide ARDI alerts to any industrial system that speaks the open OPC-UA protocol.. This allows users to connect their Agency alerts to SCADA systems and industrial historians. For those cases where systems don't natively speak OPC-UA, there are a range of hardware protocol converters that can read OPC-UA data and write it directly into PLCs or make it available using other industrial protocols, such as Modbus/TCP.

Agency OPC-UA Output

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency OPC-UA Output If enabled, Agency provides feedback on all of its detections through OPC-UA. This allows you to integrate Agency into industrial alarm and sensor networks. We suggest testing the system using an OPC-UA client such as UAExplorer.

Prometheus Endpoint

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Prometheus Endpoint Agency includes an HTML endpoint you can use to record alert data to the Prometheus historian or the Telegraf system. This means that the alerts that are being raised can be distributed to a number of open-source data storage, acquisition and analytic tools.

prometheus

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

prometheus The root Agency web API endpoint provides a list of alert statuses optimised for reading into a Prometheus database. Entry_Accumulator_Very_High_Current_Usage 0 Bridle_#2_Roller_1_Very_High_Current_Usage 0 Bridle_#2_Roller_2_Very_High_Current_Usage 0 Exit_Accumulator_Very_High_Current_Usage 0

Property Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Property Attachment A property attachment connects your Agency AI to all of the ARDI assets with all of the specified properties. ParameterMeaningExamplepropertiesA list of property names[“Temperature - Target”,“Temperature - Actual”] Other Attachments

Property Pattern

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Property Pattern The PropertyPattern algorithm is used to detect a specific pattern across one or more inputs. For example, a sudden drop, quick rise, peak or trough. Usage This algorithm is used for any number of inputs. For a simple threshold alert, you would use a single point. For more complex scenarios, you'd add additional points.

Relative Binding Format

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Relative Binding Format When you define a binding, you use relative binding format to determine where each input gets its data from. It's in the form .. For example, the electrical load on the asset Pump A would be Pump A.Load. With Attachments

root (/)

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

root (/) The root Agency web API endpoint provides a JSON-encoded list of active alerts. [{ "name": "Not At Target PMT @ Prime Oven Zone 3", "element": "Alert", "priority": "2", "agent": "Not At Target PMT", "triggered": "2024-04-14 19:09:55.505207" }]

Simple Alert Configuration

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Simple Alert Configuration As well as the full agent configuration file format (which is designed to work with a wide range of complex systems), the simple alert configuration format is used to make simple, rules-based alerts for ARDI systems. These use

Single Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Single Attachment A single attachment is the simplest type of Agency attachment. It connects your Agent to one asset, based on its name. ParameterMeaningExampleassetThe name of the asset to attach toMain Pump Other Attachments Type Property Spider

Spider Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Spider Attachment A spider attachment connects your Agency AI to all of the ARDI assets linked by a specific relationship. The assets in the relationship are processed in-order, so that you can use the relative binding format to let your AI refer to properties on neighbouring assets.

Strong AI

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Strong AI A strong AI is an AI that uses machine learning techniques to learn how to raise alerts. Unlike weak AIs, strong AIs are ideal for when you have a large number of inputs or can't easily define what actually triggers the alert. Machine learning techniques explore all possible combinations of your input data to find solutions to detecting problems that aren't obvious - or in some cases even

Strong AIs

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Strong AIs A strong AI is a Machine Learning AI. Rather than being based on a fixed set of rules, it uses pre-built Machine Learning AIs, or allows you to use the recorded history in ARDI to create your own machine learning models. Examples of a strong AI include detecting when

Styles of Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Styles of Attachment There are four standard types of attachments in Agency. Single Type Property Spider

Suppression

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Suppression There are times when systems are offline where you don't want alarms to appear. Agency allows you to specify alerts that supress other alerts - preventing them from appearing when people look at your alert list. For example, an alert that your

Type Attachment

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Type Attachment A type attachment connects your Agency AI to all of your ARDI assets with the nominated type. ParameterMeaningExampleassetThe name of the type to attach toPump Other Attachments Single Property Spider

Weak AI

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Weak AI A 'weak' AI is an artificial intelligence that works using hand-crafted rules. This is the opposite of a strong AI, which uses machine learning algorithms. Weak AIs are used when there are simple, logical comparisons and tests you can perform to raise an

Weak AIs

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Weak AIs A weak AI is an AI that doesn't include any machine learning. It's instead based on a set of rules. Examples of a weak AI include detecting when... * A value is above or below a particular threshold. * A value drifts too far away from a target

Web API Endpoint

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Web API Endpoint Agency provides details of the current alerts using a REST endpoint. Normally these are available on port 9090, although this may change if there are multiple Agency services running on a single server. The following URL endpoints are available

Agency AI Framework

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Agency AI Framework The agency system connects your live ARDI data to both weak (ie. rules-based) and strong (ie. machine-learning) AI. You can use it as a tool to create alerts from your ARDI data - anything from a simple comparison, to a cloud-based machine learning analytic.

Window

anonymous@undisclosed.example.com (Anonymous) — Thu, 18 Dec 2025 22:50:17 +0000

Window The window algorithm calculates if the input values are between a set of two limits. Usage This algorithm is often used with two inputs, but can take as many as you like. It always compares the highest value with the lowest one. Parameters