abhishekjain71/k8s-rag-agent ? reverse-engineered prompt

Build me a Kubernetes knowledge assistant that can read a folder of internal documents and answer questions in a chat style way. I want it to separate real technical info from noisy or irrelevant text, then use the best matches to generate a clear answer with the most useful sources shown back to me.

It should have a simple web chat interface, a backend API, and a document ingestion flow that can handle PDFs, HTML, TXT, DOCX, and PPTX files locally. If a question is off topic or looks like prompt injection, block it with a friendly safety message instead of answering. For technical questions, search the document store, rank the best chunks, and keep some conversation memory so follow up questions still make sense.

Please make it production ready enough to run with environment variables, and include a clean way to load documents, start the app, and test it. Look up current docs online if you need to.