Build a remote HMI without going on site

1. 01

   Record traffic on site

   Insert Falcon between PLC and HMI and call log.record to capture a real shift. Every tag read and write is stored as JSON-RPC.

2. 02

   Replay in the office

   Take the log back to the office and call log.replay. Falcon republishes the traffic to MQTT with original timing, as if the line were running.

   // Topic: falcon/line-01/rpc/req
   {
     "jsonrpc": "2.0",
     "id": "replay-01",
     "method": "log.replay",
     "params": { "log": "2026-11-03_line01_shift-a", "speed": 1.0, "loop": true }
   }

3. 03

   Hand it to an AI agent

   Give the requirements to an AI coding agent. It subscribes to the replay, maps the tags, writes the PySide6 code and runs tests, fixing errors on its own.

   Build a PySide6 window for "Line 01 Press".
   Data arrives as JSON-RPC notifications on MQTT topic
   falcon/line-01/event, method "traffic.observed", params.tags.
   - Show D100 (pressure, bar) as a large gauge-style number
   - Show M20 (motor running) as a green/grey status lamp
   - "Set pressure" spin box sends JSON-RPC "tag.write" for D100
     to falcon/line-01/rpc/req
   Use paho-mqtt in a QThread and Qt signals to update the UI.

4. 04

   Validate against the replay

   Run the generated app and watch it follow the replayed data live. Refine layout, alarms and formatting until operators are happy.

   import json, sys, uuid
   import paho.mqtt.client as mqtt
   from PySide6.QtCore import QObject, Signal
   from PySide6.QtWidgets import (QApplication, QLabel, QPushButton,
                                  QSpinBox, QVBoxLayout, QWidget)
   
   GW = "falcon/line-01"
   
   class Bus(QObject):
       tags = Signal(dict)
   
       def __init__(self):
           super().__init__()
           self.c = mqtt.Client(mqtt.CallbackAPIVersion.VERSION2)
           self.c.on_message = lambda c, u, m: self.tags.emit(
               json.loads(m.payload)["params"]["tags"])
           self.c.connect("broker.local", 1883)
           self.c.subscribe(f"{GW}/event")
           self.c.loop_start()
   
       def write(self, tag, value):
           self.c.publish(f"{GW}/rpc/req", json.dumps({
               "jsonrpc": "2.0", "id": str(uuid.uuid4()),
               "method": "tag.write", "params": {"tag": tag, "value": value}}))
   
   app = QApplication(sys.argv)
   bus, win = Bus(), QWidget()
   pressure, motor = QLabel("--"), QLabel("MOTOR")
   setpoint, apply = QSpinBox(maximum=5000), QPushButton("Set pressure")
   apply.clicked.connect(lambda: bus.write("D100", setpoint.value()))
   
   def update(tags):
       if "D100" in tags: pressure.setText(f"{tags['D100']} bar")
       if "M20" in tags:
           motor.setStyleSheet(f"background:{'#2e7d32' if tags['M20'] else '#555'}")
   
   bus.tags.connect(update)
   layout = QVBoxLayout(win)
   for w in (pressure, motor, setpoint, apply): layout.addWidget(w)
   win.setWindowTitle("Line 01 — Remote HMI"); win.show()
   sys.exit(app.exec())

5. 05

   Go live

   Point the MQTT connection at the site broker and the remote HMI shows the real line. With authorization enabled, it can intervene through tag.write.

Topics, methods and tag names in these examples are illustrative. Adapt them to the product documentation and your actual tags.