Swagger: Instant, Interactive API Docs

Video not working?

It looks like your browser may not support the H264 codec. If you're using Linux, try a different browser or try installing the gstreamer0.10-ffmpeg gstreamer0.10-plugins-good packages.

Thanks! This saves us from needing to use Flash or encode videos in multiple formats. And that let's us get back to making more videos :). But as always, please feel free to message us.

We're currently looking at something called Swagger: an open source API documentation interface. We're going talk more about it soon, but the idea is basically this: if you have an API - built in any language - and you create some configuration that describes that API in a format that Swagger understands, boom! Swagger can render this beautiful interactive documentation for you. Behind the scenes, API Platform is already preparing that configuration for Swagger.

Let's play with it! Open the POST endpoint. It says what this does and shows how the JSON should look to use it. Nice! Click "Try it out"! Let's see what's in my kitchen - some "Half-eaten blue cheese", which is still... probably ok to eat. We'll sell it for $1. What a bargain! And... Execute!

Um... what happened? Scroll down. Woh! It just made a POST request to /api/cheese_listings and sent our JSON! Our app responded with a 201 status code and... some weird-looking JSON keys: @context, @id and @type. Then it has the normal data for the new cheese listing: the auto-increment id, title, etc. Hey! We already have a working API... and this just proved it!

Close up the POST and open the GET that returns a collection of cheese listings. Try this one out too: Execute! Yep... there's our one listing... but it's not raw JSON. This extra stuff is called JSON-LD. It's just normal JSON, but with special keys - like @context - that have a specific meaning. Understanding JSON-LD is an important part of leveraging API Platform - and we'll talk more about it soon.

Anyways, to make things more interesting - go back to the POST endpoint and create a second cheese listing - a giant block of cheddar cheese... for $10. Execute! Same result: 201 status code and id 2.

Try the collection GET endpoint again. And... alright! Two results, with ids 1 and 2. And if we want to fetch just one cheese listing, we can do that with the other GET endpoint. As you can see, the id of the cheese listing that we want to fetch is part of the URL. This time, when we click to try it, cool! It gives us a box for the id. Use "2" and... Execute!

This makes a very simple GET request to /api/cheese_listings/2, which returns a 200 status code and the familiar JSON format.

Content-Type Negotiation

How cool is this! A full API "CRUD" with no work. Of course, the trick will be to customize this to our exact needs. But sheesh! This is an awesome start.

Let's try to hit our API directly - outside of Swagger - just to make sure this isn't all an elaborate trick. Copy the URL, open a new tab, paste and... hello JSON! Woh! Hello... API doc page again?

It scrolled us down to the documentation for this endpoint and executed it with id 2... which is cool... but what's going on? Do we actually have a working API or not?

Built into API Platform is something called Content-Type negotiation. Conveniently, when you execute an operation, Swagger shows you how you could make that same request using curl at the command line. And it includes one critical piece:

-H "accept: application/ld+json"

That says: make a request with an Accept header set to application/ld+json. The request is hinting to API Platform that it should return the data in this JSON-LD format. Whether you realize it or not, your browser also sends this header: as text/html... cause... it's a browser. That basically tells API Platform:

Hey! I want the CheeseListing with id 2 in HTML format.

API Platform responds by doing its best to do exactly that: it returns the HTML Swagger page with CheeseListing id 2 already showing.

Faking the Content-Type

This isn't a problem for an API client because setting an Accept header is easy. But... is there some way to kinda... "test" the endpoint in a browser? Totally! You can cheat: add .jsonld to the end of the URL.

Boom! This is our API endpoint in the JSON-LD format. I called this "cheating" because this little "trick" of adding the extension is really only meant for development. In the real world, you should set the Accept header instead, like if you were making an AJAX request from JavaScript.

And, check this out: change the extension to .json. That looks a bit more familiar!

This is a great example of the API Platform philosophy: instead of thinking about routes, controllers and responses, API Platform wants you to think about creating API resources - like CheeseListing - and then exposing that resource in a variety of different formats, like JSON-LD, normal JSON, XML or exposing it through a GraphQL interface.

Where do These Routes Come From?

Of course, as awesome as that is, if you're like me, you're probably thinking:

This is cool... but how did all these endpoints get magically added to my app?

After all, we don't normally add an annotation to an entity... and suddenly get a bunch of functional pages!

Find your terminal and run:

php bin/console debug:router

Cool! API platform is bringing in several new routes: api_entrypoint is sort of the "homepage" of our api, which, by the way, can be returned as HTML - like we've been seeing - or as JSON-LD, for a machine-readable "index" of what's included in our API. More on that later. There's also a /api/docs URL - which, for HTML is the same as going to /api, another called /api/context - more on that in a minute - and below, 5 routes for the 5 new endpoints. When we add more resources later, we'll see more routes.

When we installed the API Platform pack, its recipe added a config/routes/api_platform.yaml file.

resource: .
type: api_platform
prefix: /api

This is how API Platform magically adds the routes. It's not very interesting, but see that type: api_platform? That basically says:

Hey! I want API Platform to be able to dynamically add whatever routes it wants.

It does that by finding all the classes marked with @ApiResource - just one right now - creating 5 new routes for the 5 operations, and prefixing all the URLs with /api. If you want your API URLs to live at the root of the domain, just change this to prefix: /.

I hope you're already excited, but there is so much more going on than meets the eye! Next, let's talk about the OpenAPI spec: an industry-standard API description format that gives your API Swagger superpowers... for free. Yes, we need to talk a little bit of theory - but you will not regret it.

Leave a comment!

  • 2020-06-11 Diego Aguiar

    Hey Kasper Smithsonian

    I believe you got confused by the file name. This api_platform.yaml files should live inside config/routes not inside config/packages


  • 2020-06-11 Kasper Smithsonian

    resource: .
    type: api_platform
    prefix: /api

    in api_plataform.yaml not work for me , I have to put in route.yaml of config

  • 2019-12-18 Diego Aguiar

    Hey yankeedudlizz

    I haven't done what you are asking but I think you can do it by following this part of the documentation: https://api-platform.com/do...

    I hope it helps. Cheers!

  • 2019-12-17 yankeedudlizz


    Is there any way to redefine swagger operations description (like 'Retrieve foo collection') using an `@ApiResource` configuration?
    I tried this
    * @ApiResource(
    * collectionOperations={
    * "send": {
    * "method": "POST",
    * "path": "/email/send",
    * "swaggerContext": {
    * "summary": "test summary",
    * "description": "sends a confirmation code",
    * },
    * "controller": SendEmailConfirmation::class,
    * },
    * },
    but it doesn't work