AetherGuardAI/aetherguard-rag-security-kit — reverse-engineered prompt

Build me a small Python 3.10+ SDK for AetherGuard that lets someone plug security checks into an existing RAG app without changing their vector database or embedding model.

It should be an async client that talks to the AetherGuard service using an API key and service URL from the control portal. I want the main user flows covered. One flow for ingesting document chunks plus embeddings and getting back metadata to store with vectors. One flow for taking raw vector search results and returning verified, sanitized context that is safe to pass to an LLM. One flow to verify a specific chunk later for compliance or debugging. And one flow to check if a user is allowed to retrieve data before querying the vector store.

Please make the API feel simple, use an async context manager, include sensible timeout and retry behavior, and return clear errors for denied access, ingest failures, and connection problems. Add examples and tests so someone can copy the quick start. If you need details, check the current AetherGuard docs online.