Unlocking Build Insights: A Guide to Log Detective Integration in Packit

Overview

Log Detective, a powerful AI-driven log analysis tool, has joined forces with Packit to simplify troubleshooting of failed package builds. Starting this month, every time a Packit-triggered scratch Koji build on a dist-git pull request fails, Log Detective automatically steps in. It parses build logs, identifies the root cause, and presents actionable insights — all without any manual setup. This guide walks you through everything you need to know about this new feature: how it works behind the scenes, what you can expect from the analysis, and how to make the most of it.

Prerequisites

Before diving into Log Detective, ensure your environment meets these criteria:

• Packit Service Active: You must be using Packit to automate builds for your upstream project on dist-git. This integration is enabled by default for all Packit users — no extra configuration needed.
• Koji Builds Enabled: Your Packit configuration should include scratch Koji builds triggered on pull requests. Log Detective only analyzes failures from these builds.
• Access to Packit Dashboard: You need to be able to view the Packit dashboard (e.g., via the Packit GitHub App) where analysis results are posted.
• Fedora Messaging (Optional for Advanced Users): If you want to subscribe to automated notifications, ensure your system can receive messages from the Fedora Messaging bus.

Step-by-Step Guide

1. How the Automatic Analysis Triggers

The process is fully automated. Here’s the sequence of events:

1. Packit triggers a scratch Koji build on your dist-git pull request.
2. If the build fails, Packit sends a request to the Log Detective interface server. This lightweight containerized service acts as the bridge between Packit and Log Detective.
3. The interface server receives the build logs and other artifacts, then forwards them to Log Detective for analysis.
4. Log Detective processes the logs using its AI agent (built on the BeeAI Framework). It applies the Drain template mining algorithm to extract concise, relevant snippets — only a fraction of the original log size. This reduces token usage and speeds up analysis.
5. Once complete, the analysis result is posted to the Fedora Messaging bus, where Packit picks it up and displays it on the Pull Request dashboard.

2. Interpreting the Analysis in the Packit Dashboard

After a failed build, navigate to the Pull Request page on your repository. Look for a new section labeled Log Detective Analysis (or similar) within the Packit build results. The analysis typically includes:

• Issue Statement: A clear, concise statement of what caused the build failure (e.g., “Missing dependency: python3-foo”).
• Suggested Solution (Optional): A recommendation on how to fix the issue, based solely on the logs.
• No Extra Source Data: Remember, Log Detective does not consult external sources like package repositories or upstream documentation. Its suggestions come only from the build artifacts.

Example dashboard output:

“Build failed due to a syntax error in the spec file at line 42. Ensure the %prep section is correctly formatted with proper macro expansion.”

3. Example Scenario: A Failed Koji Build

Let’s walk through a typical case. Suppose you push a commit that modifies your .spec file to add a new %check section. The Koji build fails with a cryptic error. Without Log Detective, you might spend minutes scrolling through hundreds of log lines. With the integration:

1. Packit triggers the build; it fails.
2. Within seconds, Log Detective analyzes the log.
3. On the dashboard, you see: “The %check scriptlet attempted to run a test binary that is not installed. Consider adding BuildRequires: python3-pytest.”
4. You add the missing dependency, push a new commit, and the build succeeds.

This saves time and reduces the learning curve for new packagers.

4. No Extra Setup Required

One of the biggest advantages is that you don’t need to configure anything. Unlike the Copr integration where you click an “Ask AI” button, Packit triggers the analysis automatically on failure. You don’t choose which logs to send or tune a prompt — everything is handled server-side.

Common Mistakes and Misunderstandings

Overreliance on AI Suggestions

Log Detective uses a general-purpose large language model. While it often provides accurate insights, it can sometimes misinterpret logs—especially if the error is highly specific or involves domain knowledge not present in the training data. Always verify the suggested solution manually before applying it.

Ignoring Context Beyond Logs

The tool does not have access to the broader Fedora ecosystem, upstream repository history, or your project’s specific workflow. A suggestion like “install version X of library Y” might be outdated or conflict with other dependencies. Use the analysis as a starting point, not the final word.

Expecting Perfect Results Every Time

Log Detective is designed to assist inexperienced packagers. If you have years of experience building Fedora packages, you may find the suggestions too basic or occasionally irrelevant. Don’t be discouraged — it’s a tool to lower the barrier for newcomers.

Forgetting the Token-Saving Strategy

Because Log Detective extracts only snippets (using the Drain algorithm), some context may be lost. If you find the analysis vague or insufficient, check the full build log yourself. The snippet approach is a trade-off between speed and completeness.

Summary

Log Detective integration brings AI-powered log analysis to Packit users automatically. Key takeaways:

• No setup required: Works out of the box for all Packit-triggered Koji build failures.
• Fast and efficient: Uses snippet extraction to minimize token usage and analysis time.
• Dashboard presentation: Results appear on the Pull Request page, with an issue statement and optional fix suggestion.
• Aimed at newcomers: Experienced maintainers may find limited value, but it reduces friction for those new to packaging.
• Continuous improvement: The Log Detective team plans to expand support to other build systems and integrate additional data sources in the future.

To get help or contribute feedback, visit the Log Detective GitHub repository or the Packit project page.