InotGrumpy is a research agent we built partly to learn the stack, partly because we were tired of answers that sounded confident and turned out wrong. He doesn't just answer — he looks things up first.
Under the hood: LangGraph for the reasoning loop, SEA-LION running locally via Ollama, and Tavily for web search. The result is slower than a raw LLM, but far less likely to make things up. That's the trade-off we care about.
We named him honestly. He's not trying to be charming. He's trying to be right. View the repo →