When your Python code breaks, you get a traceback. You can see where it failed. You almost never know why it failed — what was actually in memory at that moment.
I built F. Incantatem to fix this. Add one decorator, and when your code crashes, an LLM captures the full context — stack trace, source code, actual variable values — and explains what went wrong and how to fix it.
It integrates as a decorator, CLI tool, or IPython extension. Works with OpenAI, OpenRouter, or locally via Ollama. Cautious mode automatically redacts secrets and PII before transmission. Zero core dependencies.
I'm sharing it because I keep thinking other people must be debugging these same impossible bugs. Happy to answer questions or take feedback.
Also worth noting that there's some more work to be done (see roadmap).