AI-powered Test Generation Straight from your Debugger
summary
A plugin named dert extracts live debugger state — call stack, real input and output values, and build files — and feeds that context via RAG from debugger runtime to an LLM (Gemini, also supports OpenAI and Anthropic) to generate or update concrete unit tests; a Kotlin multiplatform gallery demo produces an exact grayscale test and the plugin is released open source under MIT. A second plugin, perforator, fetches Grafana Pyroscope profiles (and uploaded Google Cloud Profiler files), renders IDE badges and highlights bottlenecks via Pyroscope API integration, but faces WebCode SDK and documentation hurdles and token/context limits when sending large stack traces to the model.
AI-written summary · about
chapters (15)
- 0:00Introduction to projects
- 0:50Unit testing and LLM context
- 2:52Leveraging debugger data
- 4:15Context gathering strategy
- 6:06Dessert plugin demonstration
- 7:53Technical challenges and insights
- 10:40Reducing LLM hallucinations
- 12:05Observability as context
- 17:10Introduction to perforator
- 18:56Profiling integration workflow
- 22:45Demo and performance optimization
- 26:14Development hurdles
- 28:51Future of IDE as context manager
- 30:55Call to action and conclusion
- 32:39Q&A session
related talks
41mAI-Augmented Testing: How Generative AI, Prompt Eng Turn Testers Into Superheroes
58mAI Assistant: Don’t Code Alone
30mDevoxxGenie: Your AI Assistant for IDEA
50mDeveloper Productivity – DIY (with LLM) or How I Learned to Stop Worrying and Love the AI
46mTDD & generative AI - a perfect pairing?
41mAI Driven Mobile Testing
26mAI and the Future of Testing
40mDavid Fogl: Leveraging Generative AI for Enhanced Software Testing