Mario Gimenez·Apr 20Your LLM is answering the same question over and over.A proxy that stores LLM responses by meaning, not exact text — built with Kotlin, Spring Boot, pgvector and ONNX Runtime.
Mario Gimenez·Mar 15Can an LLM read a distributed trace?Building an LLM-powered Root Cause Analysis agent with Hexagonal Architecture, Grafana Tempo, Gemini and Ollama — and every lesson learned.