> For the complete documentation index, see [llms.txt](https://docs.rootcause.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.rootcause.ai/more-details/digital-twin/version-history.md). # Version History A [Digital Twin](/more-details/digital-twin.md) is not a single fixed model. Every time you train it — on a new Dataset, a changed configuration, or a refreshed dataset — RootCause keeps the result as a numbered **version**. Each version bundles a configuration and the model trained from it, so you can compare versions, return to an earlier one, and see how the twin has evolved. One version is **current** at any time: the one your [simulations](/more-details/digital-twin/simulation-types.md) run against. The others stay on record, ready to be switched back to. *** ## The Version History view Open a twin and, from its Home, open the **Versions** card. The page lists every version with its status, creation time, type, and model fit, with the current version marked; click a row to see its details, what changed compared to the version it was derived from, and to switch to it.

The Version History page listing two versions: v1.1.0 as the Current trained Static Model with a 71.7% model fit, and v1.0.0 below it with 69.6% — Version History. Here v1.1.0 is Trained and Current at a 71.7% model fit, with v1.0.0 preserved below it; select a row to inspect a version or switch back to it.

Each version carries a **status**: * **Current** — the version simulations run against. Exactly one version is current at a time. * **Trained** — a completed version, ready to be switched to. * **Failed** — training did not complete. The version is kept on record but cannot be used; retrain or delete it. A version mid-training shows its progress in place until it resolves to Trained or Failed. A version that failed before causal discovery finished has no variables to report: its **Variables** and **Relationships** read **Not extracted** rather than 0, and **Not extracted yet** while a build is still running. A version that reads 0 has run discovery and genuinely found nothing. *** ## How versions are numbered Versions follow a three-part **major.minor.patch** scheme, and the kind of change you make decides which number advances: * **Major** version (e.g. 1.0.0 → 2.0.0) — the **data** changed: a different Dataset, a change in the underlying data itself, or a different set of included fields. A model built on different data is a different model, so it earns a new major number. * **Minor** version (e.g. 1.0.0 → 1.1.0) — a change to the prior knowledge (required and forbidden links) or the ordering constraints. Changing the algorithm versions, the training options or the variable roles alone advances the patch number instead. * **Patch** version (e.g. 1.0.0 → 1.0.1) — everything else. Any **Retrain model** run advances the patch number whatever you changed in the builder, and a **Re-discover & train** run that touched neither the data nor the prior knowledge and ordering does too. The major and minor rules apply only to **Re-discover & train** runs. See [Configuration for Existing Twin › What happens when you run](/more-details/digital-twin/configuration.md#what-happens-when-you-run). *** ## Creating a new version From a twin's Config tab, click **Modify model**. This opens the twin builder — the same wizard used to [create the twin in the first place](/user-guide/creating-digital-twin.md) — seeded with the current version's configuration: data, type, fields, roles, prior knowledge, training options, and algorithm versions.

The Modify model screen: a choice between Re-discover with new priors and Edit the graph & retrain, the data inherited from v1.0.0, and the domain-knowledge step seeded with the current model — Modifying a twin. The builder opens seeded with the current version's configuration; the method chosen in step 1 decides how the next version is built.

The builder's final step launches the work with one of two actions, depending on the method you chose: * **Re-discover & train** — re-run causal discovery with your updated priors, then train. Use this when the data or your prior knowledge changed. * **Retrain model** — retrain directly on your edited graph, with no discovery step. Use this when you adjusted the graph by hand. *** ## Training and the result While a version trains, the twin shows a live status banner — **Discovering relationships…**, then **Training model…** — and a list of processing stages — preparing data, causal discovery, then, when the options are enabled, confounder modeling and equation discovery, and finally building the model. The configuration cannot be edited until training finishes.

A twin mid-training: Discovering Relationships with the Causal Discovery stage in progress, the five processing stages listed ending in Building and Evaluating Digital Twin, the stage progress bars, and the Versions card reading v1.0.0 created Just now — A version in training. Each of the five stages reports progress; the run can be cancelled.

On success the banner turns to **Model trained & ready**, the version's status becomes Trained and it becomes the current one, and simulations can run against it. You can switch back to an earlier trained version from Version History. The Config tab now shows the version's configuration, and **Modify model** starts the next iteration.

A successfully trained twin: a Model trained & ready banner, the causal graph, and cards reporting v1.1.0 as Trained with a 71.7% model fit — A finished version — here v1.1.0, trained and current at a 71.7% model fit. The model fit score reports how well this version fits the data — see Reviewing Model Quality.

*** ## Seeing every version at a glance The Digital Twins list shows each twin with its training status and owner — the fastest way to scan what exists across a workspace. Per-version detail lives in each twin's **Versions** page.

The Digital Twins list showing a trained twin with its status and owner — The management list. Open a twin's Versions card for the per-version breakdown.

To compare two versions' metrics side by side, use the **Version comparison** selector on the twin's Evaluation tab — see [Model Comparison](/more-details/digital-twin/model-comparison.md). *** ## Other Working with a Digital Twin pages * [Exploring the Causal Model](/more-details/digital-twin/exploring-causal-model.md) — graph layouts and variable details. * [Reviewing Model Quality](/more-details/digital-twin/model-quality.md) — predictive accuracy and per-variable metrics. * [Inspecting Causal Relationships](/more-details/digital-twin/causal-relationships.md) — individual edges and their statistics. * [Configuration for Existing Twin](/more-details/digital-twin/configuration.md) — the fields behind every version. See [Digital Twin overview](/more-details/digital-twin.md) — general overview.