Advanced Patterns

Overview

Advanced design patterns for continuously improving production Agentic AI system performance. The Self-Improving Agent Loop provides a closed-loop architecture that combines human feedback with automated evaluation to enhance agent behavior, and ADR documents record the rationale and trade-offs for design decisions. Knowledge Feature Store covers 3-plane feature management combining ontologies and Knowledge Graphs, while Semantic Caching covers LLM Gateway-level semantic caching optimization.

Document List

📄️ Self-Improving Loop

5-stage loop design and safety mechanisms for self-hosted SLMs to autonomously learn and improve from production traces based on Karpathy's autosearch concept

📄️ ADR: Self-Improving Loop

Architecture Decision Record documenting principles, scope, responsibilities, and rollback boundaries to be agreed upon before introducing the Self-Improving Agent Loop to production

📄️ Knowledge Feature Store

3-plane design integrating ontology and Knowledge Graph into traditional Feature Store to reduce hallucinations, enable provenance tracking, and enhance domain entity utilization

📄️ Semantic Caching

LLM Gateway-level semantic caching strategy and implementation options comparison (GPTCache, Redis Semantic Cache, Portkey, Helicone, Bifrost+Redis)