Angular + LLM + RAG: My Production Stack
When you have to deliver a real product, the RAG stack changes completely from tutorials. On the frontend I use Angular 18 with standalone components and Server-Sent Events for response streaming without blocking the UI.
The backend: FastAPI + LangChain, OpenAI embeddings, Chroma as local vector store (or Pinecone in cloud), and retriever with reranking. Critical production points: cost (embedding cache), latency (precompute at ingest time), and observability with LangSmith.
Install the backend with: pip install fastapi langchain chromadb openai langsmith && npm install @angular/core@18