Handle authorization with ease in both Nuxt and Nitro.
This module does not implement ACL or RBAC. It provides low-level primitives that you can use to implement your own authorization logic.
[!NOTE] In the future, this module could be available as a Nitro module and a Nuxt module, but Nitro module are not yet ready.
To learn more about this module and which problem it solves, checkout my blog post about Authorization in Nuxt.
Install the module to your Nuxt application with one command:
npx nuxi module add nuxt-authorization
That's it! You can now use the module in your Nuxt app โจ
[!NOTE] You can take a look at the playground to see the module in action.
Before using the module and defining your first ability, you need to provide 2 resolvers. These functions are used internally to retrieve the user but you must implement them. This allows the module to be agnostic of the authentication layer.
For the Nuxt app, create a new plugin in plugins/authorization-resolver.ts
:
export default defineNuxtPlugin({
name: 'authorization-resolver',
parallel: true,
setup() {
return {
provide: {
authorization: {
resolveClientUser: () => {
// Your logic to retrieve the user from the client
},
},
},
}
},
})
This function is called every time you check for authorization on the client. It should return the user object or null
if the user is not authenticated. It can by async.
For the Nitro server, create a new plugin in server/plugins/authorization-resolver.ts
:
export default defineNitroPlugin((nitroApp) => {
nitroApp.hooks.hook('request', async (event) => {
event.context.$authorization = {
resolveServerUser: () => {
// Your logic to retrieve the user from the server
},
}
})
})
[!NOTE] Read more about the
event.context
This resolver is setup within the hook request
and receive the event. You can use it to retrieve the user from the session or the request. It should return the user object or null
if the user is not authenticated. It can by async.
Generally, you use a plugin to fetch the user when the app starts and then store it. Resolver functions should only return the stored user and not fetch it again (otherwise, you could have severe performance issues).
nuxt-auth-utils
The module nuxt-auth-utils
provides an authentication layer for Nuxt. If you use this module, you can use the following resolvers:
Nuxt plugin:
export default defineNuxtPlugin({
name: 'authorization-resolver',
parallel: true,
setup() {
return {
provide: {
authorization: {
resolveClientUser: () => useUserSession().user.value,
},
},
}
},
})
Nitro plugin:
export default defineNitroPlugin((nitroApp) => {
nitroApp.hooks.hook('request', async (event) => {
event.context.$authorization = {
resolveServerUser: async () => {
const session = await getUserSession(event)
return session.user ?? null
},
}
})
})
Easy!
[!NOTE] With the upcoming Nuxt4, you can store your abilities in the
app/utils/abilities.ts
file. Then, you can import them in your server folder using~/utils/abilities
. See an example with the project Orion.[!NOTE] With Nuxt4, a new
shared
directory will be introduced to easily share code between the client and the server. See the issue.
Now the resolvers are set up, you can define your first ability. An ability is a function that takes at least the user, and returns a boolean to indicate if the user can perform the action. It can also take additional arguments.
I recommend to create a new file utils/abilities.ts
to create your abilities:
export const listPosts = defineAbility(() => true) // Only authenticated users can list posts
export const editPost = defineAbility((user: User, post: Post) => {
return user.id === post.authorId
})
If you have many abilities, you could prefer to create a directory utils/abilities/
and create a file for each ability. Having the abilities in the utils
directory allows auto-import to work in the client while having a simple import in the server ~/utils/abilities
.
By default, guests are not allowed to perform any action and the ability is not called. This behavior can be changed per ability:
export const listPosts = defineAbility({ allowGuest: true }, (user: User | null) => true)
Now, unauthenticated users can list posts.
To use ability, you have access to 3 bouncer functions: allows
, denies
, and authorize
. Both of them are available in the client and the server. The implementation is different but the API is the same and it's entirely transparent the developer.
The allows
function returns a boolean if the user can perform the action:
if (await allows(listPosts)) {
// User can list posts
}
The denies
function returns a boolean if the user cannot perform the action:
if (await denies(editPost, post)) {
// User cannot edit the post
}
The authorize
function throws an error if the user cannot perform the action:
await authorize(editPost, post)
// User can edit the post
You can customize the error message and the status code per return value of the ability. This can be useful to return a 404 instead of a 403 to keep the user unaware of the existence of the resource.
export const editPost = defineAbility((user: User, post: Post) => {
if(user.id === post.authorId) {
return true // or allow()
}
return deny('This post does not exist', 404)
})
allow
and deny
are similar to returning true
and false
but deny
allows to return a custom message and status code for the error.
Most of the times, you API endpoints will use the authorize
. This can be the first line of the endpoint if no parameters are needed or after a database query to check if the user can access the resource. You do not need to catch the error since it's a H3Error
and will be caught by the Nitro server.
The allows
and denies
functions are useful in the client to perform conditional rendering or logic. You can also use them to have a fine-grained control on you authorization logic.
The module provides 2 components help you to conditionally show part of the UI. Imagine you have a button to edit a post, unauthorized users should not see the button.
<template>
<Can
:ability="editPost"
:args="[post]" // Optional if the ability does not take any arguments
>
<button>Edit</button>
</Can>
</template>
The Can
component will render the button only if the user can edit the post. If the user cannot edit the post, the button will not be rendered.
As a counterpart, you can use the Cannot
component to render the button only if the user cannot edit the post.
<template>
<Cannot
:ability="editPost"
:args="[post]" // Optional if the ability does not take any arguments
>
<p>You're not allowed to edit the post.</p>
</Cannot>
</template>
The Bouncer
component offers a more flexible and centralized way to handle both can and cannot scenarios within a single component. Instead of using separate Can
and Cannot
components, you can leverage the Bouncer component and its named slots to handle both states in a unified block.
<Bouncer
:ability="editPost"
:args="[post]" // Optional if the ability does not take any arguments
>
<template #can>
<button>Edit</button>
</template>
<template #cannot>
<p>You're not allowed to edit the post.</p>
</template>
</Bouncer>
This module, both code and design, is heavily inspired by the Adonis Bouncer. It's a well written package and I think reinventing the wheel every time is unnecessary.