Getting the client within workers

The River client working a job is stored in a value on the context provided to workers. The client can be retrieved from the context (or any contexts derived from it) using the ClientFromContext helper:

func (w *MyWorker) Work(ctx context.Context, job *river.Job[MyArgs]) error {
    client := river.ClientFromContext[pgx.Tx](ctx)

    ...
}

The type parameter pgx.Tx corresponds to the generic type parameter of the underlying Client[pgx.Tx]. When using the database/sql driver, the type parameter should be *sql.Tx:

client := river.ClientFromContext[*sql.Tx](ctx)

If ClientFromContext is called on a context which isn’t derived from the work context, it will panic. This situation indicates a programming error so most users will find the panic more convenient, although a non-panicking version is also available:

func (w *MyWorker) Work(ctx context.Context, job *river.Job[MyArgs]) error {
    client, err := river.ClientFromContextSafely[pgx.Tx](ctx)
    if err != nil {
        return fmt.Errorf("error getting client from context: %w", err)
    }

    ...
}

The work context exposes the client, including its plugins, hooks, and middleware:

builder.worker(myJob, context -> {
    context.client().insert(notify, new NotifyArgs(context.args().accountId()));
});

Use job.client from a running worker:

def work(job)
  job.client.insert NotifyArgs.new(account_id: job.args.fetch("account_id"))
end

WorkContext::client returns the client when the worker is running under River:

let client = context.client().expect("running River worker");
client.insert(NotifyArgs { account_id: job.args.account_id }).await?;

The worker context includes client:

workers.add(myJob, async ({ client, job }) => {
  await client.insert(notify, { accountId: job.args.accountId });
});

For River Pro customers, the riverpro.Client is also available on the context via the riverpro package:

riverproClient := riverpro.ClientFromContext[pgx.Tx](ctx)

A worker running under River::Pro::Client receives that same client through job.client, including its Pro APIs:

def work(job)
  client = job.client
  run = client.workflow(job.metadata.fetch("workflow_id"))
end

This example assumes the job is a workflow task. No separate context lookup or cast is required.