### Return a Response Directly

The most common case would be [returning a Response directly as explained later in the advanced docs](../advanced/response-directly.md){.internal-link target=_blank}.

```Python hl_lines="8  10-11"
{!> ../../../docs_src/response_model/tutorial003_02.py!}
```

// package's DefaultScheme has conversion functions installed which will unpack the
// JSON stored in RawExtension, turning it into the correct object type, and storing it
// in the Object. (TODO: In the case where the object is of an unknown type, a
// runtime.Unknown object will be created and stored.)
//
// +k8s:deepcopy-gen=true
// +protobuf=true
// +k8s:openapi-gen=true
message RawExtension {

In some cases, you might still have to use Pydantic's version of `dataclasses`. For example, if you have errors with the automatically generated API documentation.

In that case, you can simply swap the standard `dataclasses` with `pydantic.dataclasses`, which is a drop-in replacement:

```{ .python .annotate hl_lines="1  5  8-11  14-17  23-25  28" }
{!../../../docs_src/dataclasses/tutorial003.py!}
```

In this case, you could want to document how that external API *should* look like. What *path operation* it should have, what body it should expect, what response it should return, etc.

## An app with callbacks

!!! tip
    The next sections are **not necessarily "advanced"**.

    And it's possible that for your use case, the solution is in one of them.

## Read the Tutorial first

The next sections assume you already read the main [Tutorial - User Guide: Security](../../tutorial/security/index.md){.internal-link target=_blank}.

        ```

    * So, the colon in `main:app` would be equivalent to the Python `import` part in `from main import app`.

* `--workers`: The number of worker processes to use, each will run a Uvicorn worker, in this case, 4 workers.

* `--worker-class`: The Gunicorn-compatible worker class to use in the worker processes.
    * Here we pass the class that Gunicorn can import and use with:

        ```Python

import org.jetbrains.kotlin.analysis.api.descriptors.symbols.calculateHashCode
import org.jetbrains.kotlin.analysis.api.descriptors.symbols.descriptorBased.base.KtFe10DescSymbol
import org.jetbrains.kotlin.analysis.api.descriptors.symbols.descriptorBased.base.toKtType
import org.jetbrains.kotlin.analysis.api.descriptors.symbols.isEqualTo
import org.jetbrains.kotlin.analysis.api.descriptors.symbols.pointers.KtFe10NeverRestoringSymbolPointer

internal object KDocReferenceResolver {
    /**
     * [symbol] is the symbol referenced by this resolve result.
     *
     * [receiverClassReference] is an optional receiver type in
     * the case of extension function references (see [getTypeQualifiedExtensions]).
     */
    private data class ResolveResult(val symbol: KtSymbol, val receiverClassReference: KtClassLikeSymbol?)

In this case, if you had **multiple containers**, by default, when Prometheus came to **read the metrics**, it would get the ones for **a single container each time** (for the container that handled that particular request), instead of getting the **accumulated metrics** for all the replicated containers.

---

If you are using a third party library that communicates with something (a database, an API, the file system, etc.) and doesn't have support for using `await`, (this is currently the case for most database libraries), then declare your *path operation functions* as normally, with just `def`, like:

```Python hl_lines="2"
@app.get('/')
def results():
    results = some_library()
    return results
```