Skip to main content
The Monitor class collects and aggregates metrics from all orchestrator components, provides periodic health checks, and emits real-time metrics snapshots.

​
Monitor Class

​
Constructor

constructor(
  config: MonitorConfig,
  workerPool: WorkerPool,
  taskQueue: TaskQueue
)
​
config
MonitorConfig
required
Configuration for health checks and timeouts
​
workerPool
WorkerPool
required
Worker pool instance to monitor
​
taskQueue
TaskQueue
required
Task queue instance to track

​
Configuration

interface MonitorConfig {
  healthCheckInterval: number; // Polling interval in seconds
  workerTimeout: number;       // Worker timeout threshold in seconds
}

​
Methods

​
start()

Starts the monitoring loop with periodic health checks and metrics emission. 
start(): void

​
stop()

Stops the monitoring loop and clears all timers. 
stop(): void

​
getSnapshot()

Returns the current metrics snapshot. 
getSnapshot(): MetricsSnapshot
​
MetricsSnapshot
object
Complete metrics snapshot with current state

​
Callback Registration

The monitor supports various callbacks for event-driven monitoring: 
onTimeout(callback: (workerId: string, taskId: string) => void): void
onEmptyDiff(callback: (workerId: string, taskId: string) => void): void
onMetrics(callback: (snapshot: MetricsSnapshot) => void): void

​
Usage Example

import { Monitor } from "@longshot/orchestrator";

const monitor = new Monitor(
  { healthCheckInterval: 10, workerTimeout: 1800 },
  workerPool,
  taskQueue
);

// Register metrics callback
monitor.onMetrics((snapshot) => {
  console.log(`Active workers: ${snapshot.activeWorkers}`);
  console.log(`Pending tasks: ${snapshot.pendingTasks}`);
  console.log(`Tokens used: ${snapshot.totalTokensUsed}`);
});

// Register timeout callback
monitor.onTimeout((workerId, taskId) => {
  console.warn(`Worker ${workerId} timed out on task ${taskId}`);
});

monitor.start();

// Later...
monitor.stop();

​
Metrics Collection

The monitor automatically collects:
  • Task statistics: Pending, running, completed, failed counts
  • Worker statistics: Active workers, available capacity
  • Token usage: Total tokens consumed across all LLM calls
  • Cost tracking: Estimated USD cost based on token usage
  • Merge statistics: Merge attempts, successes, conflict rate
  • Finalization metrics: Build/test results from final sweep

​
Health Checks

Periodic health checks detect:
  • Worker timeouts: Workers that exceed workerTimeout threshold
  • Empty diffs: Workers that complete with no file changes
  • Suspicious tasks: Tasks with anomalous behavior patterns

Build docs developers (and LLMs) love

Get started for freeTalk to us