Show HN: I used Jev to build an off-topic detector for meetings
Category: ai-ml
Tags: meeting-tools, decision-models, speech-recognition, llm-classification
Score: 4.7/10 (Innovation: 5, Technical: 5, Documentation: 3, Utility: 4)
An off-topic detector for meetings built on Jev/Clef decision models, using browser-based speech recognition and a Cloudflare-deployed classifier to flag when conversations drift from a stated goal. It's an interesting practical demonstration of decision-model APIs as a cheaper, more reactive alternative to autoregressive LLMs for classification tasks, though the tool itself is a personal demo with untested heuristics.
Target audience: developers exploring LLM-based classification and meeting productivity tools
Repository: https://maxharrison.me/posts/offtrackometer/
View on Hacker News