Tusharyadav21/artha ? reverse-engineered prompt

Build me a local only AI knowledge base app called Artha that runs entirely on my own machine with Docker Compose. I want a web app where I can upload documents, ask questions about them, and get streamed answers in a chat UI. It should support two chat experiences, one for project documents with retrieval over uploaded files, and one for personal chat memory that keeps track of people, places, and other entities across conversations. Each chat should have its own permanent URL so refreshing the page does not lose the conversation.

Please wire it up so everything stays local, using a local LLM and local embeddings, with background processing for document parsing, chunking, indexing, and entity extraction. Include a clean backend API, a modern frontend, persistence for chats and files, and the core data stores needed for search and memory. I also want basic observability, auth and safety checks, and a simple way to run the whole thing from one command. If anything is unclear, look up the current docs online first.