SWARMTRACE — Multi-Agent Bug Investigation

SWARMTRACE is an AI-powered software investigation system designed to help developers understand why a bug happened, not simply generate a possible fix. PROBLEM: Debugging difficult software failures often requires developers to investigate several disconnected sources of evidence. They may need to inspect source code, reconstruct Git history, identify the change that introduced a regression, examine existing tests, and determine what test coverage is missing. This process is time-consuming and it is difficult to maintain a clear chain of reasoning from the original bug to the final fix. Most AI coding tools focus primarily on generating or modifying code rather than providing a transparent, evidence-driven investigation. SOLUTION: SWARMTRACE treats debugging as a structured investigation. A bug report is passed to specialized investigators that examine different evidence sources. The Code Detective analyzes the implementation, the Git Historian reconstructs relevant historical changes, and the Test Detective examines existing tests and missing regression coverage. Their findings are collected as structured evidence and synthesized into a root-cause analysis. The system then produces a concrete fix proposal, recommended regression tests, verification results, and a final investigation report. This creates an explainable chain from bug report → evidence → root cause → fix → verification. TARGET USERS: SWARMTRACE is designed primarily for software developers, engineering teams, students learning software debugging, and maintainers working with unfamiliar or complex codebases. USER INTERACTION: A developer submits or selects a bug investigation target through the SWARMTRACE interface and starts an investigation. The system coordinates the specialized investigation workflow and presents the resulting evidence, root cause, proposed fix, verification, and final report through the dashboard. CREATIVITY AND UNIQUENESS: The key idea is to treat software debugging as an evidence investigation rather than simply asking an AI to generate code. Different investigators independently examine different sources of evidence, allowing conclusions to be corroborated across code, history, and tests. The result is a more transparent debugging process where developers can understand why the system reached its conclusion. The demonstration includes a realistic authentication/session regression specifically designed to require evidence from multiple sources, making the investigation workflow observable and verifiable rather than a simple code-generation demonstration.