ShellJobManager

Category: Agent

Source: shell_job_manager.dart

Enums

JobStatus

Value	Description
running
failed

Classes

sealed JobEvent

JobStarted

Constructor

JobStarted(this.id, this.command)

Properties

Property	Type	Description
id	int
command	String

JobExited

Constructor

JobExited(this.id, this.exitCode)

Properties

Property	Type	Description
id	int
exitCode	int

JobError

Constructor

JobError(this.id, this.error)

Properties

Property	Type	Description
id	int
error	Object

ShellJob

Constructor

ShellJob({
    required this.id,
    required this.command,
    required this.startTime,
    required this.handle,
    required this.output,
    this.traceSpan,
  })

Properties

Property	Type	Description
id	int
command	String
startTime	DateTime
handle	RunningCommandHandle
output	LineRingBuffer	Interleaved stdout and stderr output, capped by the ring buffer limits.
status	JobStatus
exitCode	int?
traceSpan	ObservabilitySpan?

ShellJobManager

Manages the lifecycle of background shell jobs.

Each job's output is captured into a [LineRingBuffer] and status transitions are broadcast as [JobEvent]s, so the UI can update in real time without polling.

Constructor

ShellJobManager(this.executor, {this._obs})

Properties

Property	Type	Description
executor	CommandExecutor
events	Stream<JobEvent> get
jobs	List<ShellJob> get	All known jobs (running, exited, and killed), sorted by ID (oldest first).
events	Stream<JobEvent> get
jobs	List<ShellJob> get	All known jobs (running, exited, and killed), sorted by ID (oldest first).

Methods

Future<ShellJob> start(String command)

Starts [command] as a background job and begins capturing its output.

Returns immediately with the [ShellJob] handle. Output from both stdout and stderr is fed into the job's [LineRingBuffer], and a [JobStarted] event is emitted on [events].

ShellJob? getJob(int id)

Future<void> kill(int id)

Sends SIGTERM to the job with the given [id].

No-op if the job doesn't exist or has already exited.

Future<void> shutdown()

Tears down the manager, stopping all running jobs.

Sends SIGTERM first and waits briefly for graceful exit, then follows up with SIGKILL for any stubborn processes. The [events] stream is closed after cleanup, so no further events will be emitted.