Skip to content

Collection Structure

General Structure

On start-up, the MindsDB database consists of 2 collections: databases and predictors.

You can verify it by running the following MQL commands:

USE mindsdb;
SHOW collections;

On execution, we get:

| Collections_in_mindsdb    |
| databases                 |
| predictors                |

The predictors Collection

All the trained machine learning models are visible as new documents inside the predictors collection.

The predictors collection stores information about each model in the JSON format, as shown below.

    "name" : "model_name",
    "status" : "status",
    "accuracy" : 0.999,
    "predict" : "value_to_be_predicted",
    "update_status" : "update_status",
    "mindsdb_version" : "",
    "error" : "error_info",
    "select_data_query" : "",
    "training_options" : ""


Name Description
"name" The name of the model.
"status" Training status (generating, or training, or complete, or error).
"accuracy" The model accuracy (0.999 is a sample accuracy value).
"predict" The name of the target column to be predicted.
"update_status" Training update status (up_to_date, or updating, or available).
"mindsdb_version" The MindsDB version used while training ( is a sample version value).
"error" Error message stores a value in case of an error, otherwise, it is null.
"select_data_query" It is required for SQL API, otherwise, it is null.
"training_options" Additional training parameters.

The databases Collection

All the Mongo database connections are stored inside the databases collection, as shown below.

    "name" : "mongo_int",
    "database_type" : "mongodb",
    "host" : "",
    "port" : 27017,
    "user" : null


Name Description
"name" The name of the integration.
"database_type" The database type (here, mongodb).
"host" The Mongo host.
"port" The Mongo port.
"user" The Mongo user.