The section:

```Python
from .routers import items, users
```

means:

* Starting in the same package that this module (the file `app/main.py`) lives in (the directory `app/`)...
* look for the subpackage `routers` (the directory at `app/routers/`)...
* and from it, import the submodule `items` (the file at `app/routers/items.py`) and `users` (the file at `app/routers/users.py`)...

The same example from above, returning an `HTMLResponse`, could look like:

```Python hl_lines="2  7  19"
{!../../docs_src/custom_response/tutorial003.py!}
```

/// warning

      } else if (c1 <= Character.MAX_HIGH_SURROGATE) {
        // If the high surrogate was the last character, return its inverse
        if (index == end) {
          return -c1;
        }
        // Otherwise look for the low surrogate following it
        char c2 = seq.charAt(index);
        if (Character.isLowSurrogate(c2)) {
          return Character.toCodePoint(c1, c2);
        }
        throw new IllegalArgumentException(

        cliRequest.topDirectory = topDirectory;
        // We're very early in the process, and we don't have the container set up yet,
        // so we rely on the JDK services to eventually look up a custom RootLocator.
        // This is used to compute {@code session.rootDirectory} but all {@code project.rootDirectory}
        // properties will be computed through the RootLocator found in the container.

		log.Panicf("no source in tree for import %q (from import %s in %s): %v", name, fromPath, fromDir, err)
	}

	context := w.context
	if context == nil {
		context = &build.Default
	}

	// Look in cache.
	// If we've already done an import with the same set
	// of relevant tags, reuse the result.
	var key string
	if usePkgCache {
		if tags, ok := pkgTags[dir]; ok {
			key = tagKey(dir, context, tags)

	// already have the sizes from the central directory so we can
	// just ignore these.

	return nil
}

func readDirectoryEnd(r io.ReaderAt, size int64) (dir *directoryEnd, baseOffset int64, err error) {
	// look for directoryEndSignature in the last 1k, then in the last 65k
	var buf []byte
	var directoryEndOffset int64
	for i, bLen := range []int64{1024, 65 * 1024} {
		if bLen > size {
			bLen = size
		}

    bazel test $TFCI_BAZEL_COMMON_ARGS //tensorflow/tools/api/tests:api_compatibility_test
    echo "You have to re-generate the TF API goldens and have the API changes reviewed."
    echo "Look at the instructions for ':api_compatibility_test -- --update_goldens=True'"
}

# See b/279852433 (internal).
# TODO(b/279852433) Replace deps(//tensorflow/...) with deps(//...)

                  for (Integer node : graph.nodes()) {
                    Set<Integer> unused = graph.successors(node);
                  }
                  /*
                   * Also look up an earlier node so that, if the graph is using MapRetrievalCache,
                   * we read one of the fields declared in that class.
                   */
                  Set<Integer> unused = graph.successors(first);

Next, for an individual proxy we will check if it could possibly be impacted by the change. For example, we know a sidecar never is impacted by a `Gateway` update, and we can also look at scoping (from `Sidecar.egress.hosts`) to further restrict update scopes.

You would of course use the same ideas you read in [About FastAPI versions](versions.md){.internal-link target=_blank} to set the ranges of versions.

For example, your `requirements.txt` could look like:

```
fastapi[standard]>=0.113.0,<0.114.0
pydantic>=2.7.0,<3.0.0
```

And you would normally install those package dependencies with `pip`, for example:

<div class="termy">

```console