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.

You may invoke a CI script of your choice by following these instructions:

```bash
cd tensorflow-git-dir

    await do_stuff()
```

Quando você cria um gerenciador de contexto ou um gerenciador de contexto assíncrono como acima, o que ele faz é: antes de entrar no bloco `with`, ele executa o código antes do `yield`, e após sair do bloco `with`, ele executa o código depois do `yield`.

No nosso exemplo de código acima, não o usamos diretamente, mas passamos para o FastAPI para que ele o use.

		// FIXME: log returned error. ignore time being.
		_ = target.insertStmt.Close()
	}

	if target.db != nil {
		return target.db.Close()
	}

	return nil
}

// Executes the table creation statements.
func (target *MySQLTarget) executeStmts() error {
	_, err := target.db.Exec(fmt.Sprintf(mysqlTableExists, target.args.Table))
	if err != nil {
		createStmt := mysqlCreateNamespaceTable

	return cs.processors["delete"]
}

func (cs *callbacks) Row() *processor {
	return cs.processors["row"]
}

func (cs *callbacks) Raw() *processor {
	return cs.processors["raw"]
}

func (p *processor) Execute(db *DB) *DB {
	// call scopes
	for len(db.Statement.scopes) > 0 {
		db = db.executeScopes()
	}

	var (
		curTime           = time.Now()
		stmt              = db.Statement
		resetBuildClauses bool
	)

import org.jetbrains.kotlin.psi.psiUtil.startOffsetSkippingComments
import java.io.File
import java.nio.file.Path


abstract class IncubatingApiReportWorkAction : WorkAction<IncubatingApiReportParameter> {

    override fun execute() {
        try {
            val versionToIncubating = mutableMapOf<Version, MutableSet<IncubatingDescription>>()
            val repositoryRoot = parameters.repositoryRoot.get().asFile.toPath()

    client =
      client
        .newBuilder()
        .addNetworkInterceptor(
          Interceptor { chain: Interceptor.Chain? ->
            throw AssertionError() // Network interceptors don't execute.
          },
        ).build()
    webServer.enqueue(
      MockResponse
        .Builder()
        .webSocketUpgrade(serverListener)
        .build(),
    )

 * until [nextPart] returns null. After calling [nextPart] any preceding parts should not be read.
 *
 * Typical use loops over the parts in sequence:
 *
 * ```kotlin
 * val response: Response = call.execute()
 * val multipartReader = MultipartReader(response.body!!)
 *
 * multipartReader.use {
 *   while (true) {
 *     val part = multipartReader.nextPart() ?: break
 *     process(part.headers, part.body)
 *   }
 * }

        it.proceed(it.request())
      }.build()

  private fun makeRequest(client: OkHttpClient): Handshake {
    val call = client.newCall(Request(server.url("/")))
    return call.execute().use { it.handshake!! }
  }

* **Starlette**:
    * Tendrá el siguiente mejor rendimiento, después de Uvicorn. De hecho, Starlette usa Uvicorn para ejecutarse. Así que probablemente solo pueda ser "más lento" que Uvicorn por tener que ejecutar más código.
    * Pero te proporciona las herramientas para construir aplicaciones web sencillas, con enrutamiento basado en paths, etc.

You don't even have to check if the token exists to return an error. You can be sure that if your function is executed, it will have a `str` in that token.

You can try it already in the interactive docs:

<img src="/img/tutorial/security/image03.png">

We are not verifying the validity of the token yet, but that's a start already.