Struct ServerlessWorker 

pub struct ServerlessWorker { /* private fields */ }

Available on crate feature serverless only.

RunPod Serverless worker loop.

Implementations

impl ServerlessWorker

pub fn new(config: WorkerConfig) -> Result<Self>

Create a worker from explicit configuration.

pub fn from_env() -> Result<Self>

Create a worker from RunPod Serverless environment variables.

pub async fn run<H, Fut>(&self, handler: H) -> Result<()>

where H: Fn(WorkerJob) -> Fut + Send + Sync + 'static, Fut: Future<Output = Result<Value, String>> + Send + 'static,

Run the worker loop forever, with up to WorkerConfig::concurrency jobs in flight at once.

Spawns concurrency independent copies of the poll/process loop (concurrency defaults to 1, from RUNPOD_WORKER_CONCURRENCY via WorkerConfig::from_env) — each copy is still internally sequential (take_job → await handler to completion → post_result → repeat), but running concurrency of them side by side lets that many jobs actually overlap, which matters for any handler backed by a worker pool wider than one (e.g. multiple GPU inference slots). A single heartbeat task is still shared across all of them and reports every concurrently active job id (see start_heartbeat/active_jobs).

handler needs Send + Sync + 'static (and Fut: Send + 'static) so it can be shared across the spawned tasks — virtually always true already for realistic async worker handlers running on a multi-threaded Tokio runtime.

pub async fn run_once<H, Fut>(&self, handler: &H) -> Result<bool>

where H: Fn(WorkerJob) -> Fut, Fut: Future<Output = Result<Value, String>>,

Poll for one job and process it if present.

Returns true when a job was processed and false when RunPod returned no job for this worker.

Trait Implementations

impl Clone for ServerlessWorker

fn clone(&self) -> ServerlessWorker

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.