System boundaries and trade-offs
Architecture Guides
Follow decision-oriented guides through the components, alternatives, failure modes, and controls of production AI systems.
Guide
LLM Application Boundaries
Separate model behavior, application policy, tools, state, and validation into testable system boundaries.
Start with the core concept GuideRetrieval System Design
Trace ingestion, chunking, search, reranking, generation, citations, and evaluation as one evidence pipeline.
Start with the core concept GuideAgent Control Paths
Design tool authorization, state, memory, failure handling, and human control around probabilistic decisions.
Start with the core concept GuideReliable AI Operations
Connect evaluation, latency, throughput, cost, privacy, security, monitoring, and release decisions.
Start with the core concept