Struct PythonWorkerSupervisor 

pub struct PythonWorkerSupervisor { /* private fields */ }

One persistent worker per Python provider. Invocations are deliberately serial; callers receive a stable busy error instead of entering a queue.

Implementations

impl PythonWorkerSupervisor

pub fn cancel_active(&self) -> bool

Cooperatively marks the active invocation cancelled, then terminates its complete process tree after a short grace period. The next invocation starts a fresh worker.

impl PythonWorkerSupervisor

pub fn status(&self) -> PythonWorkerStatus

Returns a bounded, redacted snapshot without executing provider code.

impl PythonWorkerSupervisor

pub fn new( identity: PythonWorkerIdentity, interpreter: PythonInterpreter, config: PythonSupervisorConfig, ) -> Arc<Self>

pub fn new_with_capabilities( identity: PythonWorkerIdentity, interpreter: PythonInterpreter, config: PythonSupervisorConfig, capabilities: &HostCapabilities, ) -> Arc<Self>

pub async fn reset_quarantine(&self)

Clears a crash-loop quarantine after an explicit operator action.

pub fn deactivate(&self)

Stops new work while allowing an invocation that already owns the worker to complete on this generation.

pub fn activate(&self)

Re-enables a retained generation during atomic rollback.

pub fn acquire_dispatch(&self) -> bool

Reserves a call while a registry generation is still active.

pub fn release_dispatch(&self)

Releases a call reservation after the routed invocation completes.

pub async fn suspend(&self)

Permanently parks this generation after all already-routed calls drain.

pub async fn preflight(&self) -> Result<Value, PythonSupervisorError>

pub async fn invoke( &self, provider: &str, action: &str, arguments: Value, surface: ProviderSurface, snapshot_id: &str, timeout_override: Duration, ) -> Result<Value, PythonSupervisorError>

pub async fn invoke_with_context( &self, provider: &str, action: &str, arguments: Value, options: PythonInvocationOptions<'_>, ) -> Result<Value, PythonSupervisorError>

pub async fn shutdown(&self)

pub async fn drain_and_shutdown(&self)

Stop accepting work, then shut the worker down within the configured grace period. Used when a registry generation is retired.