GenAI Replay Queue Analysis

Stream Pipelines includes an AI-powered feature that provides human-readable summaries of replay queue errors and troubleshooting recommendations. GenAI is used to analyze error patterns, identify error root causes, and advise corrective actions.

Replay queue errors are analyzed by GenAI at these levels:

  • Replay Queue: Aggregated overview of pipelines with errors, highlighting dominant error types and urgency.
  • Pipeline: Analysis of categories of errors that affect a specific pipeline with the follow-up prompts for deeper investigation at these levels:
    • Error category: Root cause analysis of a specific error category within a pipeline.
    • Individual error: Detailed analysis of a single failed event.

Each analysis generates a structured report with these details:

  • Pipeline details
  • Error details
  • Expiration status
  • Payload analysis
  • Troubleshooting recommendations with confidence scores
  • Documentation references
  • Recommended follow-up actions

All GenAI summaries include a disclaimer indicating that the content is AI-generated and can contain inaccuracies. We recommend that you verify critical information before you act based on AI-generated recommendations.

If the number of errors or pipelines exceeds the analysis limit, a message prompts you to narrow down the scope, for example, by selecting a specific pipeline for the analysis.

GenAI summarization enforces tenant isolation. Only data that belongs to your tenant is analyzed. No prompts or AI responses are retained beyond the request.

Replay Queue summary

When the GenAI feature is enabled for your tenant, a short AI-generated summary is displayed in the Replay Queue page header. This automatically-generated summary provides an overview of the current error situation across all pipelines.

The header summary includes the number of pipelines with errors, the dominant error type, total failed events, and an urgency indicator for records that approach their retention expiry.

To access the detailed GenAI report, click the Arrow icon or the View full report link in the Replay Queue page header. A side panel is displayed with a complete AI-generated analysis.

Pipeline-level analysis

To analyze a specific pipeline, select Data Fabric > Replay Queue, select a pipeline from the list, and click Analyze in the panel with error categories. The GenAI side panel is displayed with a structured report for the selected pipeline, including the error category analysis, potential root causes, and recommended actions.

Drilling into details

Each GenAI summary or analysis includes suggested follow-up actions. For deeper investigation, click any follow-up prompt. For example, you can drill down from a Replay Queue summary to a specific pipeline, from a pipeline-level analysis to an error category or sample error payload, and from an error category-level analysis to individual error events or similar errors in other pipelines.

To return to the previous analysis level, click the Back button.

To close the panel, click Close.