So, you could add additional data to the automatically generated schema.

For example, you could decide to read and validate the request with your own code, without using the automatic features of FastAPI with Pydantic, but you could still want to define the request in the OpenAPI schema.

You could do that with `openapi_extra`:

```Python hl_lines="19-36  39-40"

            Any,
            Doc(
                """
                The type to use for the response.

                It could be any valid Pydantic *field* type. So, it doesn't have to
                be a Pydantic model, it could be other things, like a `list`, `dict`,
                etc.

                It will be used for:

		u, ok, err := globalIAMSys.CheckKey(r.Context(), accessKey)
		if err != nil {
			return auth.Credentials{}, false, ErrIAMNotInitialized
		}
		if !ok {
			// Credentials could be valid but disabled - return a different
			// error in such a scenario.
			if u.Credentials.Status == auth.AccountOff {
				return cred, false, ErrAccessKeyDisabled
			}
			return cred, false, ErrInvalidAccessKeyID

   * FeatureSpecificTestSuiteBuilder.suppressing()}.
   */
  /*
   * TODO(cpovirk): or we could make HOLES_FORBIDDEN a feature. Or we could declare that
   * implementations are permitted to throw IAE if a hole is requested, and we could update
   * test*Hole to permit IAE. (But might this ignore genuine bugs?) But see the TODO above
   * testLower, which could make this all unnecessary
   */
  public static Method[] getHoleMethods() {

    ```

 *  **New APIs to permit easy certificate pinning.** Be warned, certificate
    pinning is dangerous and could prevent your application from trusting your
    server!

 *  **Cache improvements.** This release fixes some severe cache problems
    including a bug where the cache could be corrupted upon certain access
    patterns. We also fixed a bug where the cache was being cleared due to a

Depending on your use case, you might prefer to use a different library, but if you asked me, I would probably suggest you try **Strawberry**.

Here's a small preview of how you could integrate Strawberry with FastAPI:

{* ../../docs_src/graphql/tutorial001.py hl[3,22,25:26] *}

   * time, then a request arrived and was immediately granted? This RateLimiter would immediately
   * forget about that past underutilization. This may result in either underutilization or
   * overflow, depending on the real world consequences of not using the expected rate.
   *
   * Past underutilization could mean that excess resources are available. Then, the RateLimiter

   * our class loader (which could be that of a dynamically loaded web application or OSGi bundle),
   * it would prevent our class loader from getting garbage collected.
   */
  static class DecoupledLoader implements FinalizerLoader {
    private static final String LOADING_ERROR =
        "Could not load Finalizer in its own class loader. Loading Finalizer in the current class "

<img src="/img/tutorial/extending-openapi/image03.png">

## Change the Theme

The same way you could set the syntax highlighting theme with the key `"syntaxHighlight.theme"` (notice that it has a dot in the middle):

{* ../../docs_src/configure_swagger_ui/tutorial002.py hl[3] *}

That configuration would change the syntax highlighting color theme:

<img src="/img/tutorial/extending-openapi/image04.png">

    //
    // All this makes me really wonder if there's any value in queueing here at all. A dispatcher
    // that simply loops through the subscribers and dispatches the event to each would actually
    // probably provide a stronger order guarantee, though that order would obviously be different
    // in some cases.

    /** Global event queue. */
    private final ConcurrentLinkedQueue<EventWithSubscriber> queue =