Enterprise chat application combining real-time messaging with retrieval-augmented generation (RAG) using Ollama or OpenAI models. User and IoT requests flow through a WAF-protected API Gateway into a FastAPI backend, which orchestrates message queuing, PostgreSQL + pgVector storage, and an embedding service feeding a vector database. The RAG pipeline retrieves contextual data and routes queries to AI models with safety guardrails, while monitoring and logging track system health across ingestion, processing, storage, and serving layers. This architecture demonstrates production-grade patterns for integrating LLMs into chat applications with enterprise security, caching, and observability. Fork and customize this diagram on Diagrams.so to adapt the RAG pipeline, swap AI model providers, or extend monitoring for your use case.