kontent-ai / sample-app-vue

A sample website written in JavaScript and Vue utilizing the Kontent.ai Delivery API to manage and retrieve content.
https://kontent-sample-app-vue.netlify.app
MIT License
24 stars 36 forks source link
content-delivery delivery-api hacktoberfest headless-cms kontent-ai kontent-ai-sample spa vue vuejs

Kontent.ai sample Vue.js single-page application

Netlify Status Live Demo Stack Overflow Discord

This is a sample website written in Typescript utilizing the Kontent.ai Delivery API to retrieve content. You can register your account for free at https://app.kontent.ai.

Application setup

  1. Install the latest version of NodeJS and npm. You can download both at https://nodejs.org/en/download/.
  2. Clone the sample application repository.
  3. Navigate to the root folder of the application in the command line.
  4. Type npm install to install required npm packages.
  5. Type npm run dev to start a development server.
  6. The application opens in your browser at http://localhost:5173.

Connecting to your sample project

On the first run of the app, you'll be presented with a configuration page. It will allow you to connect the app to your Kontent.ai sample project or create a new one. You'll also be able to start a trial and convert to a developer plan when the trial expires.

If you want to open the configuration page after the project is already connected to the app. Just open the URL http://localhost:5173/Admin/Configuration.

Alternatively, you can connect your project manually as per the chapter below.

Connecting to your project manually

If you want to change the source Kontent.ai project, follow these steps:

  1. In Kontent.ai, choose Environment settings from the app menu.
  2. Under the General settings, copy Environment ID.
  3. Create and open a .env.local file in the sample application folder.
  4. On the first line, add your Environment ID constant using the format VITE_VUE_APP_ENVIRONMENT_ID=00000000-0000-0000-0000-000000000000.
  5. Save the file.

Now, when you run the application, it will retrieve the content from your sample project. This setup has a higher priority than setting your sample project via the Configuration page.

Previewing content from your project

To preview unpublished content in the sample application, follow these steps:

  1. In Kontent.ai, choose Project settings from the app menu.
  2. Under Production environment settings, choose API keys.
  3. Copy your Environment ID and Preview API key.
  4. Create and open a .env.local file in the sample application folder.
  5. On the first line, add your Environment ID constant using the format VITE_VUE_APP_ENVIRONMENT_ID=00000000-0000-0000-0000-000000000000.
  6. On the next line, add your Preview API key using the format VUE_APP_PREVIEW_API_KEY=00000000-0000-0000-0000-000000000000.
  7. Save the file.

Now, when you now run the application, you will see all project content including the unpublished version of content items.

Content administration

  1. Navigate to https://app.kontent.ai in your browser.
  2. Sign in with your credentials.
  3. Manage content in the content administration interface of your sample project.

You can learn more about content editing at Kontent.ai Learn.

Content delivery

You can retrieve content either through the Kontent.ai Delivery SDKs or the Kontent.ai Delivery API:

For more info about the API, see the API reference.

You can find the Delivery and other SDKs at https://github.com/kontent-ai.

Data origin

This sample wants to showcase loading content from Kontent.ai as well as loading some part of the site from static JSON resources.

Basically, the content that you can't find in the Localization folder is loaded from Kontent.ai.

Model mapping and data fetching

There are two types of model mapping in this application:

content type -> DTO -> component

Content type definitions are generated from content types via Kontent.ai model generator tool. All types can be found in src/Models folder. The _project.ts contains information about the project structure such as project languages as well as other structure information like codenames about content types.

content type -> DTO -> view model -> component

Some models displayed in views might require an adjustment from content types. For example, the content type Cafe contains fields for city and street and we would like to have a model containing an address in the format city, street. You can find an example of such a view model in CafeModel.tsx located in the src/ViewModels folder. Converting Cafe into CafeModel can be done by the function located in src/Utilities/CafeListing.ts.

Data fetching

This solution fetches data using the Kontent.ai Delivery client. For more implementation detail on how to set up the client see src/Client.js. After your client is set up, you are able to deliver your content to your project. The following example showcases how to use a Kontent.ai delivery client to fetch data.

const fetchBrewer = () => {
  var query = Client.items<Brewer>()
    .type('brewer')
    .equalsFilter('url_pattern', route.params.brewerSlug as string)

  if(language) {
    query.languageParameter(language)
  }

  query
    .toPromise()
    .then(response => {
      // store data to the state variable of your component.
      brewer.value = response.data.items[0]
    }
}

Language fallbacks

To deal with content that is not available in the current language, this project uses a method called language fallbacks. It will fetch the content in the language set as a fallback language in the Kontent.ai project and redirect the website to the URL with a prefix of the given language. However, it is possible to disable language fallbacks by adding a filter of system.language to your query. For more information about getting localized content check this link.

var query = Client.items<AboutUs>().type('about_us');

if (language) {
  query
    .languageParameter(language)
    .equalsFilter('system.language', 'es-ES');
}

Handling 404

For the not found resources, prefixed 404 pages are used for both languages. As the content on one page should be in one language, this approach might help you to optimize SEO. If language is not set in the URL the application uses the last used language, which is set in cookies.

Deployment

You can use, for example, surge to deploy your app live. Check out the step-by-step guide on our blog.