Test Log Analyzer
An AI-powered failure analysis system for the OmniCore FlexPendant automated test framework at ABB Robotics. It pulls test logs from CI, structures them, and writes the root-cause report an engineer would otherwise piece together by hand.
When a test fails in CI, someone has to read the logs and work out why. This does the first pass.
It parses every run into structured JSONL, asks GPT-4o-mini to interpret the failures using the team’s own glossary, and publishes a Markdown report inside the DevOps workflow, ready for cross-team diagnostics.
- Azure DevOps CI integration
- Python modules that structure logs as JSONL
- GPT-4o-mini through Azure OpenAI
- Glossary-based contextual prompting
- Multi-run support
- Markdown reports in the DevOps workflow
- Role
- Intern
- When
- Aug – Oct 2025
- Stack
- Python, Azure DevOps, Azure OpenAI
- Now
- Run by the QA team
From log to root cause
Extract
Pull test logs from the CI pipelines of the OmniCore FlexPendant test framework in Azure DevOps.
Structure
Parse every run into structured JSONL, across as many runs as a diagnosis needs.
Interpret
Azure OpenAI’s GPT-4o-mini reads the failures, guided by a glossary of the team’s own terms.
Report
Root-cause summaries come out as Markdown reports inside the DevOps workflow.
I documented it and handed it to the QA team, who kept running it.
Root-cause summaries that used to be written by hand are now generated automatically, which cut manual log analysis effort by about 70%.