How Artekis Works: Playbooks, Context Models and Findings

Playbooks

What runs, and when

A playbook is a small resource in one of your Perspective projects. It names the analysis, the plant areas it covers, when it runs (on a schedule, when an alarm fires, or from a script), a cost budget, and where the findings go. Five ship in 1.2: Ask, Alarm Health, Anomaly Watch, Incident and Gateway Health.

Context model

What your equipment is

Each plant area has a context model: its assets, their tags with units, normal ranges and limits, their alarms, your rules, links to manuals and SOPs, and how assets relate. It is what lets a run know that a pressure tag is the inlet pressure of pump 3 and what normal looks like.

Building the model

Drafted from your tag tree in seconds

The editor drafts assets from what the gateway already holds: UDT definitions and instances, the folder tree, alarm configuration and history settings. On our acceptance run it drafted 19,400 tags into 242 assets in 3.1 seconds. You review the draft asset by asset and save what you accept.

Coverage. A score per asset for tags, units, ranges, relations, rules, documents and history, with a to-do list ordered by open findings and recent alarm activity.
Documents. PDFs and text manuals and SOPs, searched during Incident and Ask runs. Key-shaped strings are redacted before the text is stored. For example: a supervisor asks what to do when the hydraulic temperature runs high, and the answer quotes the machine manual’s page on the heat exchanger, by passage.
Rules. Your engineers’ knowledge in plain terms: a condition, its consequence and the action. An answer in Ask that proposes a rule can be saved to the model. For example: “If the furnace holds below 680 °C for more than 20 minutes, the next pour is suspect: hold it for inspection.”

The analyses

What each playbook does

Playbook	Reads	Produces
Alarm Health	The alarm journal over the window, up to 500,000 transitions a run	Nuisance, stale, duplicate, priority, flood and rate findings with proposed configuration changes, and the ISA-18.2 KPIs
Gateway Health	The last 24 hours of logs, Store and Forward, connections, devices and tag quality	What regressed since the previous run; the first run records the baseline
Incident	The 30 minutes before an alarm or line stop (adjustable), across the asset and related assets	Ranked causes: the first-out alarm, the first tag outside its normal range, an operator write, a configuration change, a communications loss, or unexplained
Anomaly Watch	Up to 31 days of tag history for the area’s measurements, with 30 days of baseline before it, split by running, stopped, changeover and product	Drift, out-of-band, stuck, rate and quality anomalies with their numbers, data-health findings about the historian itself, and for a confirmed anomaly a scored detector you can import as an alarm
Ask	Live gateway data, your documents and your named queries, through a fixed set of read-only tools	An answer with a citation for each claim; an uncited answer is marked as such

Where results go

Announced where your team works

Perspective

A views pack installed into the project you choose: findings, finding detail with the evidence and proposed change, alarm KPIs, the anomaly review queue and detector view, an incident card and the Ask panel. The views use your theme.

Email and reports

An HTML digest through a gateway email profile, filtered by severity, and a scheduled PDF Alarm Health report through the Reporting module.

Tags, MQTT and scripts

A read-only tag provider with open findings and incident causes, MQTT messages through the Artek MQTT suite, and callbacks into your project library.

The Findings view in Perspective: findings newest first with severity, status, playbook, asset and filters — The Findings view, installed into your own project and using its theme, here the demo plant’s dark theme: newest first, filtered by severity, status, playbook, asset, text and dates.

Setup

A checklist that tells you what is missing

The Status page in the gateway shows eight setup steps judged from the gateway’s real state: database tables, connection and secret, license or trial, the module, configuration, the views pack, playbooks and the optional incident trigger. Before anything is sent to a model, Show prompt on the Test run page displays exactly what the model would see, free, with nothing leaving the gateway.

Next step

Want to know what it would find on your data? Get in touch.

Tell us about your gateway and what you want to find out. A member of our team reads every message and replies.

info@artekis.ai