ai-rag-dotnet-angular

This repository provides an AI-powered Retrieval-Augmented Generation (RAG) document question-answering system. It is built with ASP.NET Core 8, PostgreSQL with pgvector, OpenAI, and an Angular chat UI. The backend ingests PDF and TXT documents, extracts text, splits documents into overlapping chunks, generates OpenAI embeddings, stores them in PostgreSQL using pgvector, performs semantic search, retrieves relevant chunks, builds grounded RAG context, and uses an OpenAI LLM to generate answers. The Angular app provides a chat interface over POST /api/chat. The system displays source documents, chunks, similarity scores, and retrieved chunk evidence, and instructs the model to refuse an answer when the information cannot be determined from the indexed documents. Persistent conversation history, streaming, authentication, OCR, and AI agents are not implemented. The architecture follows Clean Architecture with API, Application, Domain, and Infrastructure layers. The RAG pipeline includes document upload, text extraction, chunking, embedding, semantic search, context construction, grounded prompt, and OpenAI LLM answer generation with sources. The Chat API is single-turn and does not store conversation history.