Skip to content

Fluent Models

Models are the heart of Fluent. Unlike ORMs in other languages, Fluent doesn't return untyped arrays or dictionaries for queries. Instead, you query the database using models. This allows the Swift compiler to catch many errors that have burdened ORM users for ages.


This guide provides an overview of the Model protocol and its associated methods and properties. If you are just getting started, check out the guide for your database at Fluent → Getting Started.

Model is a protocol in the Fluent module. It extends the AnyModel protocol which can be used for type-erasure.


Both structs and classes can conform to Model, however you must pay special attention to Fluent's return types if you use a struct. Since Fluent works asynchronously, any mutations to a value-type (struct) model must return a new copy of the model as a future result.

Normally, you will conform your model to one of the convenience models available in your database-specific package (i.e., PostgreSQLModel). However, if you want to customize additional properties, such as the model's idKey, you will want to use the Model protocol itself.

Let's take a look at what a basic Model conformance looks like.

/// A simple user.
final class User: Model {
    /// See `Model.Database`
    typealias Database = FooDatabase

    /// See `Model.ID`
    typealias ID = Int

    /// See `Model.idKey`
    static let idKey: IDKey = \.id

    /// The unique identifier for this user.
    var id: Int?

    /// The user's full name.
    var name: String

    /// The user's current age in years.
    var age: Int

    /// Creates a new user.
    init(id: Int? = nil, name: String, age: Int) { = id = name
        self.age = age


Using final prevents your class from being sub-classed. This makes your life easier.

Associated Types

Model defines a few associated types that help Fluent create type-safe APIs for you to use. Take a look at AnyModel if you need a type-erased version with no associated types.


This type indicates to Fluent which database you intend to use with this model. Using this information, Fluent can dynamically add appropriate methods and data types to any QueryBuilders you create with this model.

final class User: Model {
    /// See `Model.Database`
    typealias Database = FooDatabase
    /// ...

It is possible to make this associated type generic by adding a generic type to your class or struct (i.e, User<T>). This is useful for cases where you are attempting to create generic extensions to Fluent, like perhaps an additive service provider.

final class User<D>: Model where D: Database {
    /// See `Model.Database`
    typealias Database = D
    /// ...

You can add further conditions to D, such as QuerySupporting or SchemaSupporting. You can also dynamically extend and conform your generic model using extension User where D: ... { }.

That said, for most cases, you should stick to using a concrete type-alias wherever possible. Fluent 3 is designed to allow you to harness the power of your database by creating a strong connection between your models and the underlying driver.


This property defines the type your model will use for its unique identifier.

final class User: Model {
    /// See `Model.ID`
    typealias ID = UUID
    /// ...

This will usually be something like Int, UUID, or String although you can theoretically use any type you like.


There are several overridable properties on Model that you can use to customize how Fluent interacts with your database.


This String will be used as a unique identifier for your model whenever Fluent needs one.

final class User: Model {
    /// See ``
    static let name = "user"
    /// ...

By default, this is the type name of your model.


Entity is a generic word used to mean either "table" or "collection", depending on which type of backend you are using for Fluent.

final class Goose: Model {
    /// See `Model.entity`
    static let entity = "geese"
    /// ...

By default, this property will be name.

ID Key

The ID key is a writeable key path that points to your model's unique identifier property.

Usually this will be a property named id (for some databases it is _id). However you can theoretically use any key you like.

final class User: Model {
    /// See `Model.ID`
    typealias ID = String

    /// See `Model.entity`
    static let idKey = \.username

    /// The user's unique username
    var username: String?

    /// ...

The idKey property must point to an optional, writeable (var) property with type matching ID.


There are several lifecycle methods on Model that you can override to hook into Fluent events.

method description throwing
willCreate Called before Fluent saves your model (for the first time) Cancels the save.
didCreate Called after Fluent saves your model (for the first time) Save completes. Query fails.
willUpdate Called before Fluent saves your model (subsequent saves) Cancels the save.
didUpdate Called after Fluent saves your model (subsequent saves) Save completes. Query fails.
willRead Called before Fluent returns your model from a fetch query. Cancels the fetch.
willDelete Called before Fluent deletes your model. Cancels the delete.

Here's an example of overriding the willUpdate(on:) method.

final class User: Model {
    /// ...

    /// See `Model.willUpdate(on:)`
    func willUpdate(on connection: Database.Connection) throws -> Future<Self> {
        /// Throws an error if the username is invalid
        try validateUsername()

        /// Return the user. No async work is being done, so we must create a future manually.
        return connection) { self }


The model offers basic CRUD method (create, read, update, delete).


This method creates a new row / item for an instance of your model in the database.

If your model does not have an ID, calls to .save(on:) will redirect to this method.

let didCreate = user.create(on: req)
print(didCreate) /// Future<User>


If you are using a value-type (struct), the instance of your model returned by .create(on:) will contain the model's new ID.


Two methods are important for reading your model from the database, find(_:on:) and query(on:).

/// Finds a user with ID == 1
let user = User.find(1, on: req)
print(user) /// Future<User?>
/// Finds all users with name == "Vapor"
let users = User.query(on: req).filter(\.name == "Vapor").all()
print(users) /// Future<[User]>


This method updates the existing row / item associated with an instance of your model in the database.

If your model already has an ID, calls to .save(on:) will redirect to this method.

/// Updates the user
let didUpdate = user.update(on: req)
print(didUpdate) /// Future<User>


This method deletes the existing row / item associated with an instance of your model from the database.

/// Deletes the user
let didDelete = user.delete(on: req)
print(didDelete) /// Future<Void>

Creation and Last Update Timestamps

Often an application requires knowing when a row / item is created or gets updated. If the model specifies fields for storing creation and/or update timestamps, Fluent will automatically set these values. The model simply specifies the name of the field and a matching definition.

static let createdAtKey: TimestampKey? = \.createdAt
static let updatedAtKey: TimestampKey? = \.updatedAt

var createdAt: Date?
var updatedAt: Date?

The presence of createdAtKey causes the associated field to be set when the object is created. This field will not be subsequently changed. The presence of updatedAtKey causes the associated field to be set when the object is created and whenever any update operation is performed. A model may define one, both, or neither.

Soft Delete

Fluent provides support for a two step deletion process. A soft delete marks an existing row / item so that it does not appear in query results by default but does not remove it from the database. This row / item may be restored at a later date or deleted entirely from the database when desired. This can be a useful feature, but it requires some thought by the application developer. Foreign key and uniqueness constraints in the database are impacted and query performance may be reduced by the presence of soft deleted rows. Restoration of soft deleted rows, particularly when they are part of a complex graph, requires care to ensure that related data in other tables is in a proper state.

Those warnings aside, Vapor makes soft delete easy to access and use. A model defines deletedAtKey to enable soft deletes in the same manner as for the creation and update timestamps.

static let deletedAtKey: TimestampKey? = \.deletedAt

var deletedAt: Date?

The delete operation is now a soft delete. The deletion timestamp is set and, if the model specifies updatedAtKey, the update timestamp is changed. Soft deleted rows will no longer appear in results by default, but can be included in queries by setting withSoftDeleted:true.

User.query(on: conn).filter().delete()

Restoring these rows is straightforward.

User.query(on: conn, withSoftDeleted:true).filter().restore()

The restore will clear the deletion timestamp and will also set the update timestamp. The restored rows will now appear in queries as normal.

Rows, both normal and soft deleted, may be permanently deleted by indicating force: true on the delete.

User.query(on: conn, withSoftDeleted:true).filter().delete(force: true)


If there is a desire to archive deleted rows, a better solution may be to create a separate table of deleted items and use the willDelete lifecycle method to copy the row into the archive table.


Model offers some convenience methods to make working with it easier.

Require ID

This method return's the models ID or throws an error.

let id = try user.requireID()