Since models are practically programs that TensorFlow executes, using untrusted
models or graphs is equivalent to running untrusted code.

If you need to run untrusted models, execute them inside a
[**sandbox**](https://developers.google.com/code-sandboxing). Memory corruptions
in TensorFlow ops can be recognized as security issues only if they are
reachable and exploitable through production-grade, benign models.

But there are specific cases where it's useful to get the `Request` object.

## Use the `Request` object directly

Let's imagine you want to get the client's IP address/host inside of your *path operation function*.

For that you need to access the request directly.

```Python hl_lines="1  7-8"
{!../../docs_src/using_request_directly/tutorial001.py!}
```

          cache-from: type=registry,ref=tensorflow/build:latest-${{ matrix.python-version }}
          cache-to: type=inline
      -
        name: Image digest

    }

    @Override
    public Optional<Long> sizeIfKnown() {
      if (file.isFile()) {
        return Optional.of(file.length());
      } else {
        return Optional.absent();
      }
    }

    @Override
    public long size() throws IOException {
      if (!file.isFile()) {
        throw new FileNotFoundException(file.toString());
      }
      return file.length();
    }

    }

    @Override
    public Optional<Long> sizeIfKnown() {
      if (file.isFile()) {
        return Optional.of(file.length());
      } else {
        return Optional.absent();
      }
    }

    @Override
    public long size() throws IOException {
      if (!file.isFile()) {
        throw new FileNotFoundException(file.toString());
      }
      return file.length();
    }

                                            key="labels.key_match_size"/></label>
                                    <div class="form-inline col-sm-9">
                                        <la:errors property="maxSize"/>
                                        <input type="number" name="maxSize" id="maxSize"

  # $TF_VER_FULL will resolve to e.g. "2.15.0-rc2". Since $TF_VER_FULL comes
  # from get_versions.sh, which must be run *after* update_version.py, FINAL_URI
  # can't be set inside the rest of the _upload envs.
  FINAL_URI="$TFCI_ARTIFACT_FINAL_GCS_URI/$TF_VER_FULL"
  gsutil -m rsync -d -r "$DOWNLOADS" "$FINAL_URI"

  # Also mirror the latest-uploaded folder to the "latest" directory.

     */
    fun sourceFileAndSourceRootFor(sourceFilePath: String): Pair<File, File> =
        sourceRoots.asSequence()
            .map { it.resolve(sourceFilePath) to it }
            .firstOrNull { it.first.isFile }
            ?: error("Source file '$sourceFilePath' not found, searched in source roots:\n  - ${sourceRoots.joinToString("\n  - ")}")

    private
    val KtFile.normalizedPath: String?

You write standard Python with types:

```Python
from datetime import date

from pydantic import BaseModel

# Declare a variable as a str
# and get editor support inside the function
def main(user_id: str):
    return user_id


# A Pydantic model
class User(BaseModel):
    id: int
    name: str
    joined: date
```

That can then be used like:

```Python

                                            key="labels.boost_document_rule_sort_order"/></label>
                                    <div class="form-inline col-sm-9">
                                        <la:errors property="sortOrder"/>
                                        <input type="number" name="sortOrder" id="sortOrder"