Scripting

QUBIQ exposes one API surface, system, in three execution contexts. The same call reads identically in Python and JavaScript, so a script moves between contexts without being rewritten.

The three contexts

Context	Language	Runs on	Use for
Pipeline Python	Python	The Python worker	Transformations inside a dataflow
Gateway scripts	Python	The gateway, server-side	Transactions, long work, anything a browser must not own
Client scripts	JavaScript	The operator's browser	Screen interaction, validation, UI logic

Choosing

Two questions decide where a piece of logic belongs.

Two rules cover nearly every case:

1. Anything transactional belongs on the gateway. A browser disconnect must not be able to strand an open transaction holding locks.
2. Anything that must not be lost belongs in a pipeline. Pipeline writes are covered by store-and-forward; script writes are not.

The system API

Namespace	Purpose
system.tag	Read and write namespace tags
system.db	SQL, named queries, stored procedures, transactions
system.hist	Historian trend queries
system.vars	Cross-pipeline shared variables
system.files	Read and write files in the workspace uploads folder
system.message	Push messages from the gateway down to browsers
system.script	Run a Gateway entry point server-side (JavaScript only)

Full signatures, per-method detail and worked examples: System API reference.

And context, in the browser

A browser script gets a second global, context — the view, its components, the live tag values already on screen, the operator's session, and the shell around the view (panels, toasts, popovers).

The division is worth learning early: system.* crosses the wire and is permission-gated; context.* does not and is not. A read of context.tag["..."] costs nothing because the screen already has the value; await system.tag.read("...") is a round trip that asks the server.

→ Context API reference

The same code, both languages

Python

result = system.tag.read("Line1/Filler/Motor1/Speed")
if result["quality"] == "Good" and result["value"] > 2800:
    system.tag.write("Line1/Filler/Motor1/SpeedSetpoint", 2500)

JavaScript

const result = await system.tag.read("Line1/Filler/Motor1/Speed");
if (result.quality === "Good" && result.value > 2800) {
  await system.tag.write("Line1/Filler/Motor1/SpeedSetpoint", 2500);
}

The method names are camelCase in both languages — deliberately, so a script reads the same wherever it runs.

Always check quality

r = system.tag.read("Line1/Tank1/Level")
if r["quality"] != "Good":
    return None            # do not act on an untrustworthy value

A Bad-quality read still returns a value — the last known one. Acting on it is how a dead sensor drives a live process. This is the single most important habit in QUBIQ scripting.

Batch, do not loop

# ❌ N round trips
values = [system.tag.read(p) for p in paths]

# ✅ one
values = system.tag.readAll(paths)

The same applies to database work: one query returning many rows beats many queries returning one.

The sandbox

Scripts run in a restricted environment. The standard set (system, math, json, datetime) is available; administrators can install additional Python packages, which land in a managed directory on the worker's path.

→ The script sandbox

The editor

The Python editor lints as you type, so syntax errors surface while you are writing rather than when the pipeline fires at 03:00.

The Problems panel New in 1.0.6 lists what the linter found — message, severity, rule id and Ln 4, Col 7 — and clicking a row goes there. The footer used to show a bare count, which told you there was a problem and then declined to say what; the only route to the actual fault was to notice a squiggle somewhere in a scrolled file and hover it. Every editor's diagnostics arrive through the same panel, whether they came from the Python linter, TypeScript or CSS.

One font size, everywhere. New in 1.0.6 The − / + control sets the size for every code editor in the Designer — project scripts, event and binding scripts, named queries, CSS and the pipeline node editor — and it persists, so turning it up to read a long script does not have to be done again in the next panel. The AI chat and the assistant panel have their own, scoped to the message text so the header and controls stay put.

Security

• Scripts run with the gateway's authority — a script can read any tag and any connection it names. Grant script-authoring permissions accordingly.
• Prefer named queries over string-built SQL. Where inline SQL is unavoidable, always parameterise.
• Do not put credentials in scripts or in variables. Use a connection's credential storage.
• Consequential actions are recorded in the audit journal.

In this section

Page	Contents
System API reference	Every namespace, method and signature
Python nodes	Scripting inside a pipeline
Gateway scripts	Server-side entry points
Client scripts	Browser-side screen logic
The script sandbox	What is available, and how to extend it