Progressive Delivery

ToggleAI's Progressive Delivery Pipeline Engine enables teams to deploy feature flags safely using automated, multi-stage rollouts. Rather than manually updating rollout percentages, you define templates that specify percentages, durations, and health thresholds. The edge engine automatically promotes stages or initiates rollbacks based on telemetry.

A pipeline template consists of a list of ordered stages. Each stage defines the traffic allocation, duration, and metrics threshold necessary to advance.

Integrate the TS SDK to ingest execution telemetry (health metrics and execution logs) directly to your active pipeline executions.

Ingest Health Metrics

Report application-specific metrics (e.g. response error count or transaction values) associated with a pipeline execution ID.

// Report error rate metric during canary rollout
await client.ingestPipelineHealthMetric("ex_canary_789", {
  metricName: "error_rate",
  value: 1.2, // error percentage
});

// Report latency metric
await client.ingestPipelineHealthMetric("ex_canary_789", {
  metricName: "latency_p99",
  value: 230, // milliseconds
});

Ingest Execution Logs

Report active warning or error logs occurring during pipeline trials to support diagnostics.

try {
  await executeNewFeatureCode();
} catch (error) {
  // Ingest logs related to the execution
  await client.ingestPipelineExecutionLogs("ex_canary_789", {
    message: "New feature threw unhandled runtime error",
    level: "error",
    stackTrace: error instanceof Error ? error.stack : String(error)
  });
}

Auto-rollback is triggered immediately if any monitored metric exceeds the defined safety threshold. Common configurations include:

• Error Rate: Roll back if overall API/client errors exceed 2% over a 5-minute window.
• Latency: Roll back if P99 latency spikes above 500ms.
• Custom Webhook alerts: Integrated with monitoring platforms (e.g., Datadog, Sentry, PagerDuty).