DEE-Bugr

DEE-bugr

...is an open-source, AI-augmented debugger that does what GDB, VSCode, and IntelliJ don't: it explains itself.

Built on the Debug Adapter Protocol, it supports Python, C/C++, JavaScript, and Java through a single unified interface. Under the hood, a real-time anomaly detector scans for null pointer dereferences, integer overflows, and out-of-bounds accesses after every step: no button press required. When something goes wrong, one click sends the full execution context to Llama 3.1 70B via Groq and returns a plain-English root cause analysis and an inline fix you can accept with a single click.

The visualizations are the part that doesn't exist anywhere else. Variable history timelines plot every value a variable ever held as an interactive graph, click any point to jump back to that exact step. The object graph visualizer renders linked lists, trees, and object references as live node diagrams. Mutation diff lets you pick any two execution steps and see exactly what changed in memory between them. Debug Cinema replays your entire session as a scrubable animation. None of these ship in any open-source debugger today.

DEE-bugr also introduces debugging primitives the ecosystem is missing: dependent breakpoints that only fire after another breakpoint has been hit, hit-count auto-disable, AI pre-run breakpoint suggestions, and a Beginner/Expert mode toggle that rewrites every panel's language for the audience in front of the screen.

The stack is Electron + React + TypeScript on the frontend, DAP adapters for each language runtime, and Groq (or Ollama for fully offline use) for inference. The entire thing is Apache-2.0 licensed and built in the open.