# Config Tab

The Config tab is where you define what goes into your Digital Twin—which data to analyze, which fields to include, and what constraints to apply. Think of it as the blueprint for causal discovery.

Getting the configuration right matters. Include too few variables and you'll miss important relationships. Include irrelevant ones and you'll slow down discovery and potentially find spurious patterns. The Config tab gives you precise control over this balance.

(SCREENSHOT: Config tab overview showing data view selector, field list, and configuration sections)

***

### Data View Selection

Every Digital Twin is built from a Data View. This section shows which Data View powers the current version and lets you change it when editing.

**Current Data View**

When not editing, you see the name of the selected Data View. This is the prepared dataset that causal discovery will analyze.

**Changing the Data View**

Click **Edit** to enter editing mode, then:

1. Select a different Data View from the dropdown
2. The available fields update to reflect the new source
3. Previous field selections are reset

Changing the Data View creates a new major version of the twin, since it represents a fundamentally different dataset.

(SCREENSHOT: Data View selector dropdown in edit mode)

***

### Available Fields

This section shows all columns from your Data View and lets you control which are included in causal discovery.

**Field List**

Each field shows:

* **Name** – The column name
* **Type** – Data type (Number, Category, Boolean, DateTime)
* **Include toggle** – Whether to include in analysis

**Including/Excluding Fields**

Toggle fields to include or exclude them:

* **Included fields** participate in causal discovery and can have relationships discovered
* **Excluded fields** are ignored entirely

**When to Exclude:**

* Unique identifiers (customer\_id, order\_id) – they don't have causal meaning
* Metadata columns (created\_at, updated\_by) – unless actually relevant
* Redundant fields highly correlated with others
* Fields you know are irrelevant to your analysis

**When to Include:**

* Potential causes and effects
* Potential confounders (variables that might influence multiple others)
* Anything you want to simulate or predict

(SCREENSHOT: Field list with type badges and include/exclude toggles)

***

### Temporal Dependencies (Time Series Models)

For temporal Digital Twins, this section defines time-based relationships between variables.

**What Are Temporal Dependencies?**

Temporal dependencies specify which variables can influence which others across time. If variable A at time t can affect variable B at time t+1, that's a temporal dependency.

**Auto-Generate Dependencies**

Click "Generate Dependencies" to let RootCause.ai automatically suggest temporal dependencies based on:

* Variable types and names
* Time series patterns in the data
* Common temporal relationships

**Manual Configuration**

You can also manually specify:

* Which variables have lagged effects on others
* The maximum lag to consider
* Specific known temporal relationships

(SCREENSHOT: Temporal dependencies editor showing variable pairs and lag settings)

***

### Fixed Subgraph (Known/Blocked Relationships)

This is where you encode domain knowledge about which relationships definitely exist or definitely don't.

**Known Relationships**

Edges that must exist in the discovered graph. Use these when:

* You have experimental evidence that A causes B
* Domain expertise tells you a relationship is certain
* You want to ensure the model captures a specific mechanism

**Adding a Known Relationship:**

1. Click "Add Known Relationship"
2. Select the source variable
3. Select the target variable
4. The relationship is added to the list

**Blocked Relationships**

Edges that cannot exist. Use these when:

* A relationship is logically impossible (effect can't cause its own cause)
* You want to test hypotheses by excluding certain paths
* Domain knowledge rules out a connection

**Adding a Blocked Relationship:**

1. Click "Add Blocked Relationship"
2. Select the two variables that cannot be connected
3. The block is added to the list

(SCREENSHOT: Fixed subgraph editor showing known and blocked relationship lists)

***

### Running Causal Discovery

After configuring, you'll run causal discovery to create a new version:

**First Version**

If this is a new twin with no versions:

1. Select a Data View
2. Configure fields and constraints
3. Click **Run Causal Discovery**

**New Version from Existing**

If you're modifying an existing twin:

1. Click **Edit** to enter edit mode
2. Make your changes
3. Click **Save and Create New Version**

**What Happens:**

1. Your configuration is saved
2. Causal discovery runs on the selected Data View
3. A new version is created with the discovered graph
4. You're redirected to the new version

(SCREENSHOT: Run Causal Discovery button and version creation progress)

***

### Version Changes

Changes create different types of versions:

| Change Type                  | Version Impact                      |
| ---------------------------- | ----------------------------------- |
| Change Data View             | Major version (e.g., 1.0.0 → 2.0.0) |
| Change temporal dependencies | Minor version (e.g., 1.0.0 → 1.1.0) |
| Change fixed subgraph        | Minor version                       |
| Change included fields       | Minor version                       |

A warning banner shows what kind of version change your edits will create.

(SCREENSHOT: Version change warning banner)

***

### Best Practices

**Start Broad, Then Narrow**

Include more variables initially. You can always exclude irrelevant ones in later versions based on what you learn.

**Encode What You Know**

If you're certain about relationships (from experiments, physics, or business logic), add them as Known. This guides discovery and improves results.

**Block Impossible Relationships**

If something definitely can't cause something else, block it. This prevents spurious findings and speeds up discovery.

**Document Your Reasoning**

When you add Known or Blocked relationships, remember why. You'll want to explain these choices later.

***

### Next Steps

After configuring and running discovery:

* Check the [Relationships Tab](https://docs.rootcause.ai/user-guide/digital-twin/tabs/relationships-tab) to see what was discovered
* Resolve any bidirectional edges if needed
* Proceed to [Evaluation Tab](https://docs.rootcause.ai/user-guide/digital-twin/tabs/evaluation-tab) to assess model quality