# MLflow

> Track ML experiments and manage models in MLflow: find the best run, log metrics and params, register model versions and move aliases such as champion. Start a workflow when a version is registered, an alias moves or a run finishes.

App · 17 actions. Web page: https://docs.lodol.com/docs/integrations/mlflow

## Actions

### Create run

Start a run in an experiment, to log metrics and params to from a workflow.

**Inputs**

- Experiment (`experiment`, text, required): The experiment. Pick one, or give its name, ID or link.
- Run name (`run_name`, text): A name for the run. Leave blank for one MLflow makes up.
- Tags (`tags`, object): Tags to set, each a name and a value, such as team = fraud.

**Outputs**

- `artifact_uri` (text): Where its files are kept
- `deleted` (true/false): True when it is in MLflow's trash
- `description` (text): The run's description
- `duration_seconds` (number): How long it ran, in seconds; empty while it runs
- `ended_at` (text): When it ended (ISO 8601, UTC); blank while it runs
- `experiment_id` (text): The ID of the experiment it is in
- `link` (text): Its page in MLflow
- `logged_model_ids` (list): The IDs of the models it logged (MLflow 3)
- `metrics` (object): Each metric's latest value, by name
- `params` (object): Its params, by name, as text
- `run_id` (text): The run's ID
- `run_name` (text): Its name
- `started_at` (text): When it started (ISO 8601, UTC)
- `status` (text): RUNNING, SCHEDULED, FINISHED, FAILED or KILLED
- `tags` (object): Its own tags, by name (MLflow's system tags left out)
- `user` (text): Who started it

**Example output**

```json
{
  "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
  "deleted": false,
  "description": "",
  "duration_seconds": 12.609,
  "ended_at": "2026-11-10T09:12:51Z",
  "experiment_id": "1",
  "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
  "logged_model_ids": [
    "m-677160f3846547079d1bccb8c1bcc987"
  ],
  "metrics": {
    "accuracy": 0.9111,
    "f1": 0.9191,
    "loss": 0.25
  },
  "params": {
    "C": "1.0",
    "max_iter": "200"
  },
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "run_name": "logreg-C1.0",
  "started_at": "2026-11-10T09:12:38Z",
  "status": "FINISHED",
  "tags": {
    "team": "ml-platform"
  },
  "user": "admin"
}
```

### Delete model alias

Remove an alias from a registered model. The version itself stays.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Alias (`alias`, text, required): The alias, such as champion: letters, digits, _ and - only.

**Outputs**

- `alias` (text): The alias
- `deleted` (true/false): True when it was removed; false when it wasn't set
- `model` (text): The registered model's name
- `version` (number): The version it pointed at; empty when it wasn't set

**Example output**

```json
{
  "alias": "challenger",
  "deleted": true,
  "model": "iris-classifier",
  "version": 4
}
```

### Delete model version

Delete a version of a registered model for good, with the aliases pointing at it. The run and its files stay.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Version (`version`, text, required): The version's number, such as 3. Pick one once the model is chosen.

**Outputs**

- `aliases_removed` (list): The aliases that pointed at it, removed with it
- `deleted` (true/false): True when it was deleted; false when there was none
- `model` (text): The registered model's name
- `version` (number): The version's number

**Example output**

```json
{
  "aliases_removed": [],
  "deleted": true,
  "model": "iris-classifier",
  "version": 2
}
```

### Delete run

Move a run to MLflow's trash, where it can be restored until the server's clean-up removes it.

**Inputs**

- Run ID (`run_id`, text, required): The run's ID (32 letters and digits, at the top of its page in MLflow), or a link to the run.

**Outputs**

- `already_deleted` (true/false): True when it was in the trash already
- `deleted` (true/false): True: the run is in MLflow's trash
- `experiment_id` (text): The ID of the experiment it was in
- `run_id` (text): The run's ID
- `run_name` (text): Its name

**Example output**

```json
{
  "already_deleted": false,
  "deleted": true,
  "experiment_id": "1",
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "run_name": "logreg-C1.0"
}
```

### End run

Mark a run finished, failed or killed, with the time it ended.

**Inputs**

- Run ID (`run_id`, text, required): The run's ID (32 letters and digits, at the top of its page in MLflow), or a link to the run.
- Status (`status`, text): How the run ended. Defaults to finished. One of: `FINISHED`, `FAILED`, `KILLED`. Default: `FINISHED`.

**Outputs**

- `artifact_uri` (text): Where its files are kept
- `deleted` (true/false): True when it is in MLflow's trash
- `description` (text): The run's description
- `duration_seconds` (number): How long it ran, in seconds; empty while it runs
- `ended_at` (text): When it ended (ISO 8601, UTC); blank while it runs
- `experiment_id` (text): The ID of the experiment it is in
- `link` (text): Its page in MLflow
- `logged_model_ids` (list): The IDs of the models it logged (MLflow 3)
- `metrics` (object): Each metric's latest value, by name
- `params` (object): Its params, by name, as text
- `run_id` (text): The run's ID
- `run_name` (text): Its name
- `started_at` (text): When it started (ISO 8601, UTC)
- `status` (text): RUNNING, SCHEDULED, FINISHED, FAILED or KILLED
- `tags` (object): Its own tags, by name (MLflow's system tags left out)
- `user` (text): Who started it

**Example output**

```json
{
  "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
  "deleted": false,
  "description": "",
  "duration_seconds": 12.609,
  "ended_at": "2026-11-10T09:12:51Z",
  "experiment_id": "1",
  "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
  "logged_model_ids": [
    "m-677160f3846547079d1bccb8c1bcc987"
  ],
  "metrics": {
    "accuracy": 0.9111,
    "f1": 0.9191,
    "loss": 0.25
  },
  "params": {
    "C": "1.0",
    "max_iter": "200"
  },
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "run_name": "logreg-C1.0",
  "started_at": "2026-11-10T09:12:38Z",
  "status": "FINISHED",
  "tags": {
    "team": "ml-platform"
  },
  "user": "admin"
}
```

### Find best run

Find the finished run in an experiment with the highest or lowest value of a metric, such as the most accurate.

**Inputs**

- Experiment (`experiment`, text, required): The experiment. Pick one, or give its name, ID or link.
- Metric (`metric`, text, required): The metric to compare, by its name in MLflow, such as accuracy.
- Best is (`goal`, text): Whether the highest or the lowest value is best. Defaults to highest. One of: `highest` (Highest (accuracy, F1)), `lowest` (Lowest (loss, error)). Default: `highest`.
- Filter (`filter`, text): Only runs matching this, written as in MLflow's search box, such as metrics.accuracy > 0.9 and params.model = 'rf'. Text values go in single quotes. Leave blank for every run.

**Outputs**

- `experiment_id` (text): The experiment's ID
- `experiment_name` (text): Its name
- `found` (true/false): True when a finished run has the metric
- `goal` (text): highest or lowest
- `metric` (text): The metric compared
- `run` (object): The best run; empty when none was found
- `value` (number): The best run's value of the metric; empty when none was found

**Example output**

```json
{
  "experiment_id": "1",
  "experiment_name": "iris-classifier",
  "found": true,
  "goal": "highest",
  "metric": "accuracy",
  "run": {
    "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
    "deleted": false,
    "description": "",
    "duration_seconds": 12.609,
    "ended_at": "2026-11-10T09:12:51Z",
    "experiment_id": "1",
    "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
    "logged_model_ids": [
      "m-677160f3846547079d1bccb8c1bcc987"
    ],
    "metrics": {
      "accuracy": 0.9111,
      "f1": 0.9191,
      "loss": 0.25
    },
    "params": {
      "C": "1.0",
      "max_iter": "200"
    },
    "run_id": "8c019ddca7894c08800a84f71c75318e",
    "run_name": "logreg-C1.0",
    "started_at": "2026-11-10T09:12:38Z",
    "status": "FINISHED",
    "tags": {
      "team": "ml-platform"
    },
    "user": "admin"
  },
  "value": 0.9111
}
```

### Find or create experiment

Find the experiment with a name, or create it when there is none, to start runs in.

**Inputs**

- Experiment name (`experiment_name`, text, required): The experiment's name, such as churn-model. On Databricks it is a workspace path, such as /Users/you@example.com/churn-model.
- Tags (`tags`, object): Tags for the experiment if it's created, each a name and a value.

**Outputs**

- `artifact_location` (text): Where its runs' files are kept
- `created` (true/false): True when this step made it; false when it was there already
- `created_at` (text): When it was made (ISO 8601, UTC)
- `experiment_id` (text): The experiment's ID
- `lifecycle_stage` (text): active, or deleted when it is in the trash
- `link` (text): Its page in MLflow
- `name` (text): Its name
- `tags` (object): Its own tags, by name
- `updated_at` (text): When it last changed (ISO 8601, UTC)

**Example output**

```json
{
  "artifact_location": "mlflow-artifacts:/1",
  "created": false,
  "created_at": "2026-11-10T09:12:30Z",
  "experiment_id": "1",
  "lifecycle_stage": "active",
  "link": "https://mlflow.example.com/#/experiments/1",
  "name": "iris-classifier",
  "tags": {
    "team": "ml-platform"
  },
  "updated_at": "2026-11-10T09:12:30Z"
}
```

### Get model version

Get a model version by its number or by an alias, such as the one champion points at.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Version (`version`, text): The version's number. Leave blank to give an alias instead.
- Alias (`alias`, text): An alias, such as champion, used when no version is given.

**Outputs**

- `aliases` (list): The aliases pointing at it, such as champion
- `created_at` (text): When it was registered (ISO 8601, UTC)
- `created_by` (text): Who registered it, when MLflow says
- `description` (text): Its description
- `found` (true/false): True: the version was found
- `link` (text): Its page in MLflow
- `model` (text): The registered model's name
- `model_id` (text): The logged model it was registered from (MLflow 3)
- `run_id` (text): The run it was registered from
- `source` (text): Where its files are
- `stage` (text): Its stage (legacy): None, Staging, Production or Archived
- `status` (text): READY, PENDING_REGISTRATION or FAILED_REGISTRATION
- `tags` (object): Its tags, by name
- `updated_at` (text): When it last changed (ISO 8601, UTC)
- `version` (number): The version's number

**Example output**

```json
{
  "aliases": [
    "champion"
  ],
  "created_at": "2026-11-10T09:12:58Z",
  "created_by": "",
  "description": "Logistic regression, C=1.0",
  "found": true,
  "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
  "model": "iris-classifier",
  "model_id": "m-677160f3846547079d1bccb8c1bcc987",
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
  "stage": "None",
  "status": "READY",
  "tags": {
    "validated": "true"
  },
  "updated_at": "2026-11-10T09:13:40Z",
  "version": 3
}
```

### Get run

Get a run with its status, metrics, params, tags and the models it logged.

**Inputs**

- Run ID (`run_id`, text, required): The run's ID (32 letters and digits, at the top of its page in MLflow), or a link to the run.

**Outputs**

- `artifact_uri` (text): Where its files are kept
- `deleted` (true/false): True when it is in MLflow's trash
- `description` (text): The run's description
- `duration_seconds` (number): How long it ran, in seconds; empty while it runs
- `ended_at` (text): When it ended (ISO 8601, UTC); blank while it runs
- `experiment_id` (text): The ID of the experiment it is in
- `link` (text): Its page in MLflow
- `logged_model_ids` (list): The IDs of the models it logged (MLflow 3)
- `logged_models` (list): The models the run logged
- `metrics` (object): Each metric's latest value, by name
- `params` (object): Its params, by name, as text
- `run_id` (text): The run's ID
- `run_name` (text): Its name
- `started_at` (text): When it started (ISO 8601, UTC)
- `status` (text): RUNNING, SCHEDULED, FINISHED, FAILED or KILLED
- `tags` (object): Its own tags, by name (MLflow's system tags left out)
- `user` (text): Who started it

**Example output**

```json
{
  "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
  "deleted": false,
  "description": "",
  "duration_seconds": 12.609,
  "ended_at": "2026-11-10T09:12:51Z",
  "experiment_id": "1",
  "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
  "logged_model_ids": [
    "m-677160f3846547079d1bccb8c1bcc987"
  ],
  "logged_models": [
    {
      "model_id": "m-677160f3846547079d1bccb8c1bcc987",
      "name": "model"
    }
  ],
  "metrics": {
    "accuracy": 0.9111,
    "f1": 0.9191,
    "loss": 0.25
  },
  "params": {
    "C": "1.0",
    "max_iter": "200"
  },
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "run_name": "logreg-C1.0",
  "started_at": "2026-11-10T09:12:38Z",
  "status": "FINISHED",
  "tags": {
    "team": "ml-platform"
  },
  "user": "admin"
}
```

### List registered models

List registered models with their aliases and newest version.

**Inputs**

- Name contains (`name_contains`, text): Only models whose names contain this. Leave blank for all.
- How many (`limit`, number): The most models to return, up to 200. Defaults to 25. From 1 to 200.

**Outputs**

- `count` (number): How many models are returned
- `models` (list): The registered models, by name

**Example output**

```json
{
  "count": 1,
  "models": [
    {
      "aliases": [
        {
          "alias": "champion",
          "version": 3
        }
      ],
      "created_at": "2026-11-10T09:12:56Z",
      "description": "Classifies iris flowers.",
      "latest_version": 3,
      "latest_version_stage": "None",
      "link": "https://mlflow.example.com/#/models/iris-classifier",
      "name": "iris-classifier",
      "tags": {},
      "updated_at": "2026-11-10T09:13:40Z"
    }
  ]
}
```

### Log metrics, params and tags

Log metrics, params and tags to a run in one go, such as an evaluation's scores.

**Inputs**

- Run ID (`run_id`, text, required): The run's ID (32 letters and digits, at the top of its page in MLflow), or a link to the run.
- Metrics (`metrics`, object): Metrics to log, each a name and a number, such as accuracy = 0.93.
- Params (`params`, object): Params to log, each a name and a value. MLflow won't change a param already logged.
- Tags (`tags`, object): Tags to set, each a name and a value, such as team = fraud.
- Step (`step`, number): The step the metrics are for, such as an epoch. Defaults to 0. At least 0.

**Outputs**

- `metrics_logged` (number): How many metrics were logged
- `params_logged` (number): How many params were logged
- `run` (object): The run as it is now
- `run_id` (text): The run's ID
- `step` (number): The step the metrics were logged at
- `tags_logged` (number): How many tags were set

**Example output**

```json
{
  "metrics_logged": 2,
  "params_logged": 2,
  "run": {
    "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
    "deleted": false,
    "description": "",
    "duration_seconds": 12.609,
    "ended_at": "2026-11-10T09:12:51Z",
    "experiment_id": "1",
    "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
    "logged_model_ids": [
      "m-677160f3846547079d1bccb8c1bcc987"
    ],
    "metrics": {
      "accuracy": 0.9111,
      "f1": 0.9191,
      "loss": 0.25
    },
    "params": {
      "C": "1.0",
      "max_iter": "200"
    },
    "run_id": "8c019ddca7894c08800a84f71c75318e",
    "run_name": "logreg-C1.0",
    "started_at": "2026-11-10T09:12:38Z",
    "status": "FINISHED",
    "tags": {
      "team": "ml-platform"
    },
    "user": "admin"
  },
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "step": 0,
  "tags_logged": 1
}
```

### Move model version to stage (legacy)

Move a model version to Staging, Production, Archived or None. Stages are deprecated in MLflow: use aliases where you can.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Version (`version`, text, required): The version's number, such as 3. Pick one once the model is chosen.
- Stage (`stage`, text, required): The stage to move the version to. One of: `Staging`, `Production`, `Archived`, `None`.
- Archive the versions there (`archive_existing_versions`, true/false): Archive the model's other versions in that stage. Only for Staging and Production.

**Outputs**

- `aliases` (list): The aliases pointing at it, such as champion
- `created_at` (text): When it was registered (ISO 8601, UTC)
- `created_by` (text): Who registered it, when MLflow says
- `description` (text): Its description
- `link` (text): Its page in MLflow
- `model` (text): The registered model's name
- `model_id` (text): The logged model it was registered from (MLflow 3)
- `run_id` (text): The run it was registered from
- `source` (text): Where its files are
- `stage` (text): Its stage (legacy): None, Staging, Production or Archived
- `status` (text): READY, PENDING_REGISTRATION or FAILED_REGISTRATION
- `tags` (object): Its tags, by name
- `updated_at` (text): When it last changed (ISO 8601, UTC)
- `version` (number): The version's number

**Example output**

```json
{
  "aliases": [
    "champion"
  ],
  "created_at": "2026-11-10T09:12:58Z",
  "created_by": "",
  "description": "Logistic regression, C=1.0",
  "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
  "model": "iris-classifier",
  "model_id": "m-677160f3846547079d1bccb8c1bcc987",
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
  "stage": "None",
  "status": "READY",
  "tags": {
    "validated": "true"
  },
  "updated_at": "2026-11-10T09:13:40Z",
  "version": 3
}
```

### Register model version

Register the model a run logged as a new version of a registered model, registering the model's name first if it's new.

**Inputs**

- Registered model (`model`, text, required): The registered model to add a version to. Pick one, or type a name to register a new one.
- Run ID (`run_id`, text, required): The run's ID (32 letters and digits, at the top of its page in MLflow), or a link to the run.
- Model in the run (`logged_model`, text): The model's name in the run, as logged (such as model). Leave blank for the run's only model, or the one named model.
- Description (`description`, text): A description for the new version.
- Tags (`tags`, object): Tags to set, each a name and a value, such as team = fraud.

**Outputs**

- `aliases` (list): The aliases pointing at it, such as champion
- `created_at` (text): When it was registered (ISO 8601, UTC)
- `created_by` (text): Who registered it, when MLflow says
- `created_model` (true/false): True when this step also registered the model's name
- `description` (text): Its description
- `link` (text): Its page in MLflow
- `model` (text): The registered model's name
- `model_id` (text): The logged model it was registered from (MLflow 3)
- `run_id` (text): The run it was registered from
- `source` (text): Where its files are
- `stage` (text): Its stage (legacy): None, Staging, Production or Archived
- `status` (text): READY, PENDING_REGISTRATION or FAILED_REGISTRATION
- `tags` (object): Its tags, by name
- `updated_at` (text): When it last changed (ISO 8601, UTC)
- `version` (number): The version's number

**Example output**

```json
{
  "aliases": [],
  "created_at": "2026-11-10T09:12:58Z",
  "created_by": "",
  "created_model": false,
  "description": "Logistic regression, C=1.0",
  "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
  "model": "iris-classifier",
  "model_id": "m-677160f3846547079d1bccb8c1bcc987",
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
  "stage": "None",
  "status": "READY",
  "tags": {
    "validated": "true"
  },
  "updated_at": "2026-11-10T09:13:40Z",
  "version": 3
}
```

### Search model versions

List a registered model's versions, or those registered from a run, newest first, with their aliases.

**Inputs**

- Registered model (`model`, text): Only this model's versions. Pick one, or type its name. Leave blank for every model.
- Run ID (`run_id`, text): Only versions registered from this run. Leave blank for any.
- How many (`limit`, number): The most versions to return, up to 200. Defaults to 25. From 1 to 200.

**Outputs**

- `count` (number): How many versions are returned
- `versions` (list): The versions, newest first

**Example output**

```json
{
  "count": 1,
  "versions": [
    {
      "aliases": [
        "champion"
      ],
      "created_at": "2026-11-10T09:12:58Z",
      "created_by": "",
      "description": "Logistic regression, C=1.0",
      "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
      "model": "iris-classifier",
      "model_id": "m-677160f3846547079d1bccb8c1bcc987",
      "run_id": "8c019ddca7894c08800a84f71c75318e",
      "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
      "stage": "None",
      "status": "READY",
      "tags": {
        "validated": "true"
      },
      "updated_at": "2026-11-10T09:13:40Z",
      "version": 3
    }
  ]
}
```

### Search runs

Find runs in an experiment by a metric, param or tag filter, in the order asked, with their metrics, params and tags.

**Inputs**

- Experiment (`experiment`, text, required): The experiment. Pick one, or give its name, ID or link.
- Filter (`filter`, text): Only runs matching this, written as in MLflow's search box, such as metrics.accuracy > 0.9 and params.model = 'rf'. Text values go in single quotes. Leave blank for every run.
- Order by (`order_by`, text): How to order the runs, such as metrics.accuracy DESC; separate several with commas. Leave blank for newest first.
- Status (`status`, text): Only runs with this status. Leave blank for any. One of: `FINISHED`, `FAILED`, `RUNNING`, `KILLED`.
- How many (`limit`, number): The most runs to return, up to 200. Defaults to 25. From 1 to 200.

**Outputs**

- `count` (number): How many runs are returned
- `experiment_id` (text): The experiment's ID
- `experiment_name` (text): Its name
- `runs` (list): The runs found

**Example output**

```json
{
  "count": 1,
  "experiment_id": "1",
  "experiment_name": "iris-classifier",
  "runs": [
    {
      "artifact_uri": "mlflow-artifacts:/1/8c019ddca7894c08800a84f71c75318e/artifacts",
      "deleted": false,
      "description": "",
      "duration_seconds": 12.609,
      "ended_at": "2026-11-10T09:12:51Z",
      "experiment_id": "1",
      "link": "https://mlflow.example.com/#/experiments/1/runs/8c019ddca7894c08800a84f71c75318e",
      "logged_model_ids": [
        "m-677160f3846547079d1bccb8c1bcc987"
      ],
      "metrics": {
        "accuracy": 0.9111,
        "f1": 0.9191,
        "loss": 0.25
      },
      "params": {
        "C": "1.0",
        "max_iter": "200"
      },
      "run_id": "8c019ddca7894c08800a84f71c75318e",
      "run_name": "logreg-C1.0",
      "started_at": "2026-11-10T09:12:38Z",
      "status": "FINISHED",
      "tags": {
        "team": "ml-platform"
      },
      "user": "admin"
    }
  ]
}
```

### Set model alias

Point an alias, such as champion, at a model version, moving it from the version it pointed at.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Alias (`alias`, text, required): The alias, such as champion: letters, digits, _ and - only.
- Version (`version`, text, required): The version's number, such as 3. Pick one once the model is chosen.

**Outputs**

- `alias` (text): The alias
- `model` (text): The registered model's name
- `model_version` (object): The version it points at now
- `moved` (true/false): True when it now points at a different version
- `previous_version` (number): The version it pointed at before; empty when it is new
- `version` (number): The version it points at now

**Example output**

```json
{
  "alias": "champion",
  "model": "iris-classifier",
  "model_version": {
    "aliases": [
      "champion"
    ],
    "created_at": "2026-11-10T09:12:58Z",
    "created_by": "",
    "description": "Logistic regression, C=1.0",
    "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
    "model": "iris-classifier",
    "model_id": "m-677160f3846547079d1bccb8c1bcc987",
    "run_id": "8c019ddca7894c08800a84f71c75318e",
    "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
    "stage": "None",
    "status": "READY",
    "tags": {
      "validated": "true"
    },
    "updated_at": "2026-11-10T09:13:40Z",
    "version": 3
  },
  "moved": true,
  "previous_version": 2,
  "version": 3
}
```

### Update model version

Set a model version's description, and add or change its tags, such as validated = true.

**Inputs**

- Registered model (`model`, text, required): The registered model. Pick one, or type its name.
- Version (`version`, text, required): The version's number, such as 3. Pick one once the model is chosen.
- Description (`description`, text): The version's new description. Leave blank to keep it.
- Tags (`tags`, object): Tags to set, each a name and a value, such as team = fraud.

**Outputs**

- `aliases` (list): The aliases pointing at it, such as champion
- `created_at` (text): When it was registered (ISO 8601, UTC)
- `created_by` (text): Who registered it, when MLflow says
- `description` (text): Its description
- `link` (text): Its page in MLflow
- `model` (text): The registered model's name
- `model_id` (text): The logged model it was registered from (MLflow 3)
- `run_id` (text): The run it was registered from
- `source` (text): Where its files are
- `stage` (text): Its stage (legacy): None, Staging, Production or Archived
- `status` (text): READY, PENDING_REGISTRATION or FAILED_REGISTRATION
- `tags` (object): Its tags, by name
- `updated_at` (text): When it last changed (ISO 8601, UTC)
- `version` (number): The version's number

**Example output**

```json
{
  "aliases": [
    "champion"
  ],
  "created_at": "2026-11-10T09:12:58Z",
  "created_by": "",
  "description": "Logistic regression, C=1.0",
  "link": "https://mlflow.example.com/#/models/iris-classifier/versions/3",
  "model": "iris-classifier",
  "model_id": "m-677160f3846547079d1bccb8c1bcc987",
  "run_id": "8c019ddca7894c08800a84f71c75318e",
  "source": "models:/m-677160f3846547079d1bccb8c1bcc987",
  "stage": "None",
  "status": "READY",
  "tags": {
    "validated": "true"
  },
  "updated_at": "2026-11-10T09:13:40Z",
  "version": 3
}
```
