Reference architectures
AI architecture diagrams you can start from.
Ten common AI architectures, drawn in DuctTape.io. Each one is a published diagram: open it, zoom in, export it as PNG or SVG.
RAG architecture
Retrieval-augmented generation in its basic form: ingest documents, retrieve context, prompt the model.
Advanced RAG with hybrid search and reranking
Dense and keyword retrieval combined, then a reranker that keeps only the best passages.
AI agent with MCP servers
An agent that reaches its tools through the Model Context Protocol, one server per system.
Multi-agent system: orchestrator and workers
An orchestrator plans the work and delegates to specialised agents; a human signs off.
LLM gateway with model fallback
One gateway in front of every model: cache, routing, fallback and cost tracing.
Guardrails and human-in-the-loop
Checks before and after the model, and a human approval in front of risky actions.
LLM observability and evaluation pipeline
Production traces feed online evals and a dataset; offline evals gate prompt changes.
Customer support chatbot architecture
Answers from the help center, looks up orders, and hands over to a human when needed.
Voice agent architecture
Speech-to-text, an agent with tools, and text-to-speech in one low-latency loop.
Document processing pipeline with LLMs
From inbox to structured records: OCR, LLM extraction, validation, human review.