Automated Build Failure Analysis: Log Detective Joins Packit

Introduction

Starting this month, Packit users gain access to a powerful new debugging assistant: Log Detective. This tool automatically analyzes failed scratch Koji builds triggered by dist-git pull requests, providing clear explanations and potential fixes directly within the Packit dashboard. The integration eliminates manual log inspection, streamlining the development workflow for Fedora package maintainers.

Automated Build Failure Analysis: Log Detective Joins Packit — Source: fedoramagazine.org

How It Works

Log Detective builds on the existing Copr feature where users can request AI-driven analysis via an Ask AI button. In Packit, the process is fully automated: whenever a build failure occurs, a request for analysis is sent without any user intervention. The system handles log selection, prompt tuning, and result delivery, requiring zero configuration from the developer.

Log Parsing and Analysis

Starting with version 4.0, Log Detective operates as an intelligent agent built on the BeeAI Framework. It receives all build artifacts — logs, configuration files, and error outputs — as part of the analysis request. To keep processing efficient, the tool relies on the Drain template mining algorithm and other techniques to extract concise, relevant snippets from the massive log files. These snippets represent only a small fraction of the original data, drastically reducing token usage and analysis time. By limiting irrelevant context, even relatively small language models can deliver accurate insights quickly.

Communication Architecture

When Packit detects a failed Koji build, it continues its normal error handling but now also dispatches an analysis request to a lightweight Log Detective interface server. This server acts as a bridge between the two services: it accepts the request, runs the analysis asynchronously, and posts the results onto the Fedora Messaging bus. Packit then collects the finished analysis from the bus and surfaces it in the dashboard alongside the corresponding pull request.

Result Presentation

The analysis output includes a clear statement of what went wrong during the build — for example, a missing dependency, a compilation error, or a test failure — and, when possible, a suggested solution. In its current configuration, Log Detective uses only build logs; it does not consult external sources like package history or upstream issue trackers. The results are linked directly to the triggering PR on the Packit dashboard, making it easy for maintainers to review and act on the recommendations.

Purpose and Limitations

Log Detective is a general-purpose tool with no access to broader Fedora ecosystem knowledge. It is not a substitute for the deep expertise of seasoned package maintainers; if you have years of experience, you may find little new information. Instead, it is designed to lower the barrier for newcomers who are still building familiarity with the build system. By offering automated explanations and workarounds, Log Detective helps less experienced contributors resolve issues faster and learn from common failure patterns.

Future Development

The Log Detective team continues to refine the analysis engine and expand its capabilities. Planned improvements include better integration with external data sources, support for additional build types, and improved explanation quality. This first release marks a significant step toward making automated debugging an integral part of the Packit experience, and we welcome community feedback to guide its evolution.

Tags: