Skip to content

Simulation Agent Configuration

This tutorial explains how to configure the Simulation Agent for a specific simulation source (e.g. country:demo:sim) based on the Learning session simulation agent walkthrough.

It is written for product managers, developers, and simulation engineers who need to enable AI‑powered querying of simulation data in Asset Copilot.

1. Conceptual Overview

Before configuring anything, it is important to understand the three core building blocks involved.

1.1 Main Components

Simulation Source

A data source representing a running or always‑on simulation (e.g. country:demo:sim).

Simulation Agent

An agent capable of querying simulated values, metadata, and (optionally) historical trends.

Orchestrator Agent

The agent that interprets user intent, delegates work to the simulation agent, and presents results.

Note

The Simulation Agent does not work globally by default — it is enabled per simulation source via traits.

2. Prerequisites

Ensure the following before proceeding:

  • You have admin or configuration access to the Admin / Reasoning Ops portal
  • A simulation is running (Always‑On or Planning)
  • The Simulation Agent is available in the Agent Catalog
  • The Common Simulation Meta Source exists (used for cross‑simulation metadata)

3. Step‑by‑Step Configuration

Step 1: Open Reasoning / Agent Configuration Portal

Navigate to the admin portal and open Reasoning Ops.

Reasoning Ops entry point

Step 2: Verify Simulation Agent Exists

  1. Open Agents
  2. Locate Simulation Agent
  3. Confirm:
    • Agent is enabled
    • Agent has access to simulation query tools

Simulation Agent configuration

Key fields to verify:

Field Expected Value
Preferred LLM e.g. gpt‑5.2 or codex variant
Tool bindings Include Simulation Query tools

Step 3: Identify the Simulation Source

Navigate to Sources and locate your simulation source, for example:

  • country:demo:sim
  • country:demo:testSim

This source represents the simulation whose data the agent will query.

Simulation source selected

Step 4: Add the Simulation Reasoning Trait

Critical Step

This step explicitly allows the Simulation Agent to access this simulation.

On the simulation source:

  1. Open Traits

  2. Add the trait:

    reasoning:tool:source:simulation

  3. Save / Update the source

Adding reasoning simulation trait

Step 5: Ensure Common Simulation Meta Source Is Present

The Simulation Agent often requires metadata lookup (modules, types, tag classification).

Verify there is a Common Simulation Meta Source with trait:

reasoning:tool:source:simulationMeta

Common Simulation Meta source

Info

This enables queries like:

  • "List all compressors"
  • "How many pumps are in the model?"

Step 6: (Optional) Remove Trait From Other Sources

To avoid ambiguity during testing:

  • Remove reasoning:tool:source:simulation from other simulation sources

This ensures the agent resolves queries against one simulation only.

4. Validate the Configuration

Step 7: Open Simulation Dashboard

  1. Open the simulation via Portal → Dashboard
  2. Ensure the Chat / Copilot icon is visible

Simulation dashboard with chat enabled

Step 8: Test Core Queries

Try the following queries:

What is the simulated temperature in 23VA001?

List active simulations

What is the pressure in 20PT2003?

Expected behavior:

  • Simulation Agent is invoked
  • Values resolved from the active simulation
  • Units normalized (e.g. K → °C)

Successful simulation query result

5. (Advanced) Skills for Better Interpretation

Module Retrieval Skill

To improve classification of equipment (compressors, pumps, separators):

  1. Open Skill Manager
  2. Locate Module Retrieval skill
  3. Add your simulation source to the allowed scope

Open Skill Manager

Module Retrieval skill updated

This enables queries such as:

How many compressors are in the simulation?

6. Common Failure Modes & Fixes

Symptom Likely Cause Fix
Tag not found Wrong simulation scoped Re‑check trait placement
Agent answers but no data Missing simulationMeta source Add meta source
Only snapshot, no trend Historian not enabled Enable timeseries access
Wrong tool invoked Skill missing scope Update skill configuration

7. Final Architecture Summary

  • Traits define access
  • Skills improve interpretation
  • The Simulation Agent remains intentionally narrow

8. Next Steps

Suggested follow‑ups:

  • Add historian access for trends & charts
  • Introduce control actions (start/stop/setpoint agents)
  • Validate behavior across multiple simulations

Success

You now have a fully configured Simulation Agent bound to a specific simulation source.