Automated Build Failure Diagnosis: Integrating Log Detective with Packit

Overview

Log Detective brings automated analysis to failed Koji builds triggered by Packit on dist-git pull requests. This integration eliminates manual log sifting—when a build fails, Packit automatically sends logs to Log Detective, which extracts meaningful snippets using the Drain algorithm and generates a concise diagnostic statement with a suggested solution. The result appears in the Packit dashboard, linking failure explanations directly to the triggering PR. No configuration is required; the service handles everything from log selection to AI inference.

Automated Build Failure Diagnosis: Integrating Log Detective with Packit — Source: fedoramagazine.org

Prerequisites

Before you can benefit from Log Detective, ensure your environment meets these requirements:

Packit Service: You must have Packit configured for your dist-git repository, with pull‑request triggers enabled for scratch Koji builds.
Koji Build Failure: The integration only activates when a Koji build fails. Successful builds are not analyzed.
Fedora Messaging Bus: Packit and the Log Detective interface server communicate via Fedora Messaging infrastructure―this is already set up by the service, not by you.
No Additional Credentials: The analysis uses build logs only; you do not need to provide API keys or AI model access.

Step-by-Step Guide

1. Confirm Packit Integration

Your dist‑git repository must have a packit.yaml file that defines a job for pulling requests, for example:

jobs:
- job: copr_build
  trigger: pull_request
  targets:
    - fedora-rawhide-x86_64

This ensures that when a PR is opened, Packit triggers a scratch Koji build. If you already use such a configuration, you’re ready.

2. Trigger a Scratch Koji Build

Open a pull request against your dist‑git repository. Packit will automatically start a scratch Koji build. The build status appears in the PR’s packit‑tab on the Fedora Pagure interface (or on GitLab/GitHub if mirrored).

3. Build Failure – Automatic Analysis

If the Koji build fails, Packit immediately sends a request to the Log Detective interface server. Inside the server, the Log Detective agent (built on the BeeAI Framework) downloads the full build log and other artifacts. It uses the Drain template mining algorithm to extract small but critical snippets from the logs—typically less than 1 % of the original file size. By reducing useless context, the analysis runs quickly and uses fewer tokens, even with relatively small AI models.

4. View Analysis Results

After processing, the interface server posts the analysis result (a short statement of what went wrong and optionally a solution suggestion) on the Fedora Messaging bus. Packit collects this message and displays the result in its dashboard, linked to the PR that triggered the build. You can see the explanation directly in the Packit dashboard; no additional clicks are needed to request the analysis.

5. Interpret the Output

The diagnostic typically includes two parts:

Issue statement: e.g., “Build failed due to missing dependency ‘libfoo-devel’.”
Suggested solution: e.g., “Add libfoo-devel to BuildRequires in your spec file.”

Since Log Detective uses only build logs as input, it cannot check external repositories or RPM quirks. Use the suggestion as a starting point, but verify against your own package context.

Common Mistakes and Limitations

Expecting Instant Results

While the analysis is automated, it is not instantaneous. The time depends on log size, current load on the Log Detective service, and the complexity of the failure. Check the Packit dashboard after a few minutes, not seconds.

Over‑Relying on Suggestions

Log Detective uses a general‑purpose AI model without access to your repository history, spec file details, or the Fedora package maintainer guide. Its suggestions may be generic or incomplete. For experienced maintainers the analysis often adds little value; it is designed to help newcomers who lack deep Fedora packaging experience treat it as a helpful hint, not a definitive fix.

Misunderstanding the Scope

Only failed scratch Koji builds on dist‑git pull requests are analyzed. The integration does not cover:

Successful builds.
Builds from direct pushes or manual Koji commands.
Copr builds—Copr has a separate “Ask AI” button, but that is not part of this Packit integration.

Forgetting to Check the Dashboard

The analysis result appears only in the Packit dashboard, not directly in the PR comment or email. Bookmark your Packit dashboard after setting up the integration so you don’t miss the diagnostic.

Summary

Log Detective seamlessly integrates with Packit to automatically analyze failed scratch Koji builds from dist‑git pull requests. No setup is required: Packit sends the logs, Log Detective extracts key snippets using the Drain algorithm, and a short diagnostic is posted to the Packit dashboard. While the AI’s suggestions are limited by its reliance solely on build logs, the feature dramatically lowers the barrier for newcomers to diagnose packaging failures. Check your dashboard after a failed build to see the explanation, but always validate suggestions against your package’s specific requirements.