## Prerequisites

- Prometheus and MinIO configured as explained in [document here](https://github.com/minio/minio/blob/master/docs/metrics/prometheus/README.md).
- Grafana installed as explained [here](https://grafana.com/grafana/download).

## MinIO Grafana Dashboard

flavor.

-   Snapshot API Docs: [guava][guava-snapshot-api-docs]
-   Snapshot API Diffs: [guava][guava-snapshot-api-diffs]

## Learn about Guava

-   Our users' guide, [Guava Explained]
-   [A nice collection](https://www.tfnico.com/presentations/google-guava) of
    other helpful links

## Links

-   [GitHub project](https://github.com/google/guava)

## How to Test Your Changes to TensorFlow

You may check how your changes will affect TensorFlow by:

1. Creating a PR and observing the presubmit test results
2. Running the CI scripts locally, as explained below
3. **Google employees only**: Google employees can use an internal-only tool
called "MLCI" that makes testing more convenient: it can execute any full CI job
against a pending change. Search for "MLCI" internally to find it.

It will have a property `scopes` with a list containing all the scopes required by itself and all the dependencies that use this as a sub-dependency. That means, all the "dependants"... this might sound confusing, it is explained again later below.

The `security_scopes` object (of class `SecurityScopes`) also provides a `scope_str` attribute with a single string, containing those scopes separated by spaces (we are going to use it).

Because of that, it's now recommended to instead use the `lifespan` as explained above.

## Technical Details

Just a technical detail for the curious nerds. 🤓

This is because by default, FastAPI will inspect every item inside and make sure it is serializable as JSON, using the same [JSON Compatible Encoder](../tutorial/encoder.md){.internal-link target=_blank} explained in the tutorial. This is what allows you to return **arbitrary objects**, for example database models.

!!! tip
    For new code and whenever possible, use `Annotated` as explained above. There are multiple advantages (explained below) and no disadvantages. 🍰

This is how you would use `Query()` as the default value of your function parameter, setting the parameter `max_length` to 50:

=== "Python 3.10+"

### 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!}
```

In those cases, you would probably want to build a **Docker image from scratch** as [explained above](#dockerfile), installing your dependencies, and running **a single Uvicorn process** instead of running something like Gunicorn with Uvicorn workers.

### Load Balancer

And then a new session will be created for the next request.

For that, we will create a new dependency with `yield`, as explained before in the section about [Dependencies with `yield`](dependencies/dependencies-with-yield.md){.internal-link target=_blank}.