Show HN: Semloop – Semantic loop detection for AI agent traces
Category: ai-ml
Tags: ai-agents, loop-detection, evaluation, llm-judge, deepeval
Score: 6.0/10 (Innovation: 6, Technical: 5, Documentation: 8, Utility: 5)
Semloop replaces lexical stagnation detection in AI agent traces with a semantic judge that asks whether the agent actually made progress. It targets five documented blind spots in deepeval's metric and benchmarks both an oracle and a real Jev model, showing improved detection while acknowledging remaining limitations.
Target audience: AI engineers and researchers evaluating agent traces
Repository: https://github.com/VBS2004/semloop · Python · MIT · 1 stars
View on Hacker News