RolandOS1 als RAG-System

Prompt

Act as a principal software architect. You have no filesystem access and no prior context. Design **RolandOS1 from scratch**; assume no code, data, services, or prior work exist. Produce a complete implementation plan, not code. RolandOS1 is an evidence-based personal second brain and **RAG system**. It ingests documents, performs OCR/text extraction, versioning, structure-aware chunking, embeddings, hybrid retrieval, reranking, and grounded AI answers with exact document/page citations. Updates, deletions, access control, freshness, contradiction detection, and RAG evaluation are required. Answers without sufficient evidence must say so. Required areas and ports: - HERO 80/6100: user interaction, search, upload, overview, questions. - OPERATIONS 6200: backoffice and Schaltzentrale. - LCARS 6210: visual content. - TASKS 6220: technical work. - ADMIN 6230: documents, rules, imports, settings. For each area provide separate ports/editors (WYSIWYG) start, window/dashboard editor, full menu editor, API Connection editor (single api, connected api and joined api in with api explanation), page editor. Use a central port registry; give each area a color and list all links under OPERATIONS → Bauebene. Enforce strict server-side area isolation, independent editing/restart/publishing, and controlled cross-area approval. Include governance with staging, provenance, schema/type validation, constraints, duplicate detection, review/blocker states, audit history, and no silent skipping. Design data models, APIs/events, security, UX, deployment, monitoring, backup/restore, test strategy, RAG metrics, llm-local, phased roadmap, granular backlog, acceptance criteria, risks, assumptions, and owner decisions. Deliver requirements traceability and prioritize one complete end-to-end vertical slice before expansion. German UI; English code and identifiers.