This can be very useful for setting up **resources** that you need to use for the whole app, and that are **shared** among requests, and/or that you need to **clean up** afterwards. For example, a database connection pool, or loading a shared machine learning model.

## Use Case

Let's start with an example **use case** and then see how to solve it with this.

Let's imagine that you have some **machine learning models** that you want to use to handle requests. 🤖

* Then check that the tests **pass** after the PR. ✅

* Many PRs don't have tests, you can **remind** them to add tests, or you can even **suggest** some tests yourself. That's one of the things that consume most time and you can help a lot with that.

* Then also comment what you tried, that way I'll know that you checked it. 🤓

## Create a Pull Request

    ```

The last `CommonQueryParams`, in:

```Python
... Depends(CommonQueryParams)
```

...is what **FastAPI** will actually use to know what is the dependency.

From it is that FastAPI will extract the declared parameters and that is what FastAPI will actually call.

---

In this case, the first `CommonQueryParams`, in:

=== "Python 3.8+"

    ```Python

Then, in that case, it could be simpler to have **one container** with **multiple processes**, and a local tool (e.g. a Prometheus exporter) on the same container collecting Prometheus metrics for all the internal processes and exposing those metrics on that single container.

---

When fetching certificates, ztunnel will authenticate to the CA with its own identity, but request the identity of another workload.
Critically, the CA must enforce that the ztunnel has permission to request that identity.
Requests for identities not running on the node are rejected.
This is critical to ensure that a compromised node does not compromise the entire mesh.

#### Technical Details about ORM mode

SQLAlchemy and many others are by default "lazy loading".

That means, for example, that they don't fetch the data for relationships from the database unless you try to access the attribute that would contain that data.

For example, accessing the attribute `items`:

```Python
current_user.items
```

Routes are declared in a single place, using functions declared in other places (instead of using decorators that can be placed right on top of the function that handles the endpoint). This is closer to how Django does it than to how Flask (and Starlette) does it. It separates in the code things that are relatively tightly coupled.

In the output, you can see that it shows the **PID** (process ID) of each process (it's just a number).

You can see that:

* The Gunicorn **process manager** starts with PID `19499` (in your case it will be a different number).
* Then it starts `Listening at: http://0.0.0.0:80`.
* Then it detects that it has to use the worker class at `uvicorn.workers.UvicornWorker`.

      if (obj == this) {
        return true;
      }
      if (obj instanceof Wrapper) {
        Wrapper<?> that = (Wrapper<?>) obj; // note: not necessarily a Wrapper<T>

        if (this.equivalence.equals(that.equivalence)) {
          /*
           * We'll accept that as sufficient "proof" that either equivalence should be able to
           * handle either reference, so it's safe to circumvent compile-time type checking.

So you wait for your crush to finish the story (finish the current work ⏯ / task being processed 🤓), smile gently and say that you are going for the burgers ⏸.

Then you go to the counter 🔀, to the initial task that is now finished ⏯, pick the burgers, say thanks and take them to the table. That finishes that step / task of interaction with the counter ⏹. That in turn, creates a new task, of "eating burgers" 🔀 ⏯, but the previous one of "getting burgers" is finished ⏹.