Engineering

Smarter Models Still Hallucinate: Evidence Gates for Production RAG Agents

A smarter model is still a fallible model. Better reasoning can improve average answer quality, but it does not turn a probabilistic generator into a factual database, nor does it…

  • LLM Grounding
  • Retrieval-Augmented Generation
  • Evidence Gates
  • Agent Runtime
  • Production Evaluation
Engineering

Config Enhancer Semantic Guardrails

Date: 2026-07-23 Scope: services/agent Status: implemented locally; not deployed

  • ConfigEnhancer
  • prompt grounding
  • skill passthrough
  • timer semantics
  • config cache
  • Mongo migrations