Cheap + fast enough for batch-ish work; watch rate limits; confirm Vertex vs Gemini API. Connections part_of Large Language Model
