Aviator for Analysis

Note: The Aviator integration is supported from version 25.3 as a private Beta release.

Aviator for Analysis is an AI-powered virtual assistant that helps performance engineers investigate issues, troubleshoot runtime errors, and gain insights using natural language queries.

Overview

Aviator uses generative AI to support performance analysis in the dashboard. It helps users:

  • Summarize trends and anomalies

  • Correlate metrics

  • Identify root causes of errors

  • Help new team members get started

  • Reduce the time spent diagnosing issues

  • Generate dashboard widgets based on suggestions

Supported areas and actions

Supported Details
Areas

Aviator is currently supported in the following areas of the user interface:

  • Dashboard

  • Errors tab

Actions

Aviator can answer questions related to both general and specific aspects of test runs. Supported topics include:

  • Transactions

  • Analysis

  • Anomalies

  • Metric correlations

  • Vusers

  • TRT (Transaction Response Time)

  • Response rate

  • Throughput

  • Hits

  • Hosts

  • Scripts

  • Errors

  • Other performance-related terms

Note: The following types of queries are not supported:

  • Entity generation, such as creating widgets, grids, or tabs.

  • Procedural or documentation-style questions, including instructional content.

AI usage reporting

AI query usage is tracked at the tenant level and reported through the Core Performance Engineering Platform. The AI Usage report provides aggregated visibility into AI query consumption across supported products.

For details on viewing AI usage and exporting usage reports, see AI Usage report in the OpenText Core Performance Engineering Help Center.

Prerequisites

To use Aviator for Analysis, you must have the following:

  • A Core Performance Engineering Aviator subscription. For details, contact your account team to participate in the beta program.

  • A test run must be streamed to OpenText Core Performance Engineering Analysis from each performance engineering solution. This provides the data source for performance-related questions and requests. For details, see Stream performance test results.

Use Aviator

Send a request or question about an item to Aviator. Aviator uses context from the item, such as the description fields, comments, and images, to generate relevant responses.

To use Aviator:

  1. Open the Dashboard or the Errors tab.

  2. In the banner, click the Aviator button Aviator to open the Aviator pane.

  3. Enter a request or question in the chat box, or click one of the suggested prompts.

    For guidance on using Aviator, see Usage tips and example questions.

  4. After Aviator responds, you can do the following.

    Button Action
    Copy Copy the response.
    Good response

    Leave positive feedback on the response. This feedback helps improve Smart Assistant's responses.
    Bad response

    Leave negative feedback on the response. This feedback helps improve Smart Assistant's responses.
    Copy code Copy code Copy the code (part of the received response).

  5. Continue the conversation by entering text in the chat box, or click + New topic to begin a new conversation.

    For details on using prompts, see Manage prompts.

Suggest dashboard widgets

Use Aviator to generate dashboard widgets based on suggestions or by using a predefined prompt.

To generate dashboard widgets using Aviator:

  1. In the Runs page, select a run from the Runs List and open the Dashboard tab.

  2. In the banner, click the Aviator button to open the Aviator pane.

  3. In the Aviator pane, you can:

    • Enter a request or question in the chat box to describe the widget you want to create.

    • Click the Show me the heaviest transactions prompt. This adds a widget that displays the average response times of the five heaviest transactions. This widget is generated using advanced filtering and sorting logic (for example, "heaviest" equals sort + filter). It supports all break-bys and filters, including advanced ones not shown in the UI, and multi-metric widgets.

Note:  

  • You can only add new widgets using this flow; editing existing widgets is not supported.

  • The standard widget limits, such as the maximum number of widgets per tab, still apply.

  • Measurement constraints may also affect widget generation.

Learn more about errors

You can use Aviator to learn more about errors by accessing error definitions and error logs.

To use Aviator to troubleshoot errors:

  1. In the Runs page, select a run to analyze.

  2. Click the Errors tab, or access it directly from the Dashboard using the Errors Details link.

  3. Click the Aviator button in the banner to ask about overall error patterns, such as:

    • “What are the most frequent errors in this run?”

    • “Summarize root causes for all errors.”

  4. To investigate a specific error, click the Aviator icon Aviator next to an error in the log and ask targeted questions like:

    • “What does error -26630 mean?”

    • “What caused this error in script X?”

Usage tips and example questions

The following tips can help you use Aviator effectively.

Tip Details
Start with predefined prompts

Use predefined prompts to get started with common use cases.

For details on using prompts, see Manage prompts.
Use natural language

Ask your question in plain English. Example: "What caused the errors at peak load?"

Focus on a single question

Short, focused queries often get better results. Long questions are more likely to fail.
Try follow-up questions

Build on Aviator's answers. Example: “Which scripts were most affected?” or “When did it start?”.

The following are example questions that you can ask.

Page Subject / Request
Dashboard

  • Correlate metrics:

    "Correlate throughput with Vusers over time."

  • Experiment with filters:

    Ask Aviator to filter by script, host, transaction, or error code. For example, “Identify HPS spikes on host A only.”
Errors

  • Ask about error root causes:

    “Identify the key findings and the possible root causes for the most frequent errors in this test run."

  • Drill into error patterns:

    “List common causes for -26630 errors.”

Manage prompts

You can edit, reuse, and retry prompts using the following actions.

Action Description
Edit last prompt

You can edit the last prompt sent without retyping or reconstructing complex queries.

Note:

  • Editing is only available for the last message in each chat session and becomes accessible after a page refresh or navigation between tabs or pages. Editing is not supported for older prompts.

  • Only the last edited prompt is retained; previous versions are not stored.

  • Edited messages are not marked as edited.
Reuse a prompt You can reuse any previously sent prompt in the same chat session by copying and pasting it into other contexts or chats.
Retry last prompt

You can retry the last prompt for any prompt that was stopped, halted, or failed.

Note: This option is not available for successful prompts or prompts stopped manually by the user.

See also