Web Analytics Made Easy - Statcounter
Skip to content

Custom Pipelines

Custom pipelines let teams transform raw telemetry into operational action. A pipeline can enrich events, summarize traces, run evals, detect anomalies, route alerts, create reports, or trigger remediation workflows.

What Pipelines Do

Pipelines process data on a schedule or in response to events. They can read from traces, evals, logs, dashboards, process units, external files, or custom ingestion streams.

When To Use Them

Use pipelines for:

  • Scheduled quality and cost summaries.
  • Alert enrichment and deduplication.
  • Business-process health checks.
  • Root-cause reports.
  • Data cleanup and field normalization.
  • Automated remediation through webhooks.

Prerequisites

  1. Choose the input data source.
  2. Define the output: summary, alert, dashboard field, report, webhook, or new dataset.
  3. Decide the schedule and owner.

Step-By-Step Setup

  1. Open Pipelines.
  2. Create a new pipeline.
  3. Select input sources and filters.
  4. Add steps: enrich, transform, evaluate, summarize, route, or remediate.
  5. Set schedule or event trigger.
  6. Test on a sample window.
  7. Enable and monitor run history.

Example Configuration

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
name: daily-agent-quality-summary
schedule: "0 8 * * *"
input:
  source: traces
  workflow: support-agent
steps:
  - eval: answer_quality
  - summarize: failures_by_root_cause
  - route:
      channel: slack:#ai-ops

What Appears In The Console

The pipeline list shows status, run history, duration, schedule, logs, and actions. Each run links to input records, generated outputs, alerts, or reports.

Common Mistakes

  • Scheduling expensive pipelines too frequently.
  • Running transformations without keeping source trace links.
  • Creating remediation steps before testing alert precision.
  • Not assigning an owner for failed pipeline runs.