uc-cdis/gen3-ai ? reverse-engineered prompt

Build me a Gen3 AI backend monorepo that can run a few related AI services from one repo. The main thing I need working is the embeddings service, where users can create collections and embeddings, upload them in bulk, search for similar items, and have access controlled by Gen3 tokens and row level security. Also set up the structure for the other AI services that are still in progress, like inference and a model repository, so they fit into the same pattern and can be extended later.

Use Python with a clean API first approach, good automatic API docs, and PostgreSQL for anything that needs storage. Keep shared code in one common library, and make it easy to run, test, lint, and build individual services from one place. I want a simple local setup with environment files, a single container build option for deployment, and clear README style instructions so someone can get started without digging around too much. If anything is unclear, look up current docs online if you need to.