Show HN: TrackLLM: Are the LLM APIs you rely on stable?

Category: observability

Tags: llm-observability, api-monitoring, change-detection, research-backed, log-probabilities

Score: 6.3/10 (Innovation: 7, Technical: 6, Documentation: 4, Utility: 6)

TrackLLM monitors the stability of LLM APIs by tracking log probabilities and detecting changes in model behavior over time. It is backed by two recent research papers (ICLR 2026 and ICML 2026) and aims to provide observability into API drift and reliability. The project is interesting for its academic rigor and focus on a growing operational concern for AI-dependent applications.

Target audience: AI engineers, MLOps teams, backend developers integrating LLM APIs

Repository: https://www.trackllm.net/

View on Hacker News