The cloud budget is a real constraint, not an excuse
I need alternatives that fit my cloud resource budget, and that limit is what makes me ask the price before I ask about the architecture.
My knowledge of indexing and vector search is still shallow, and it is not for lack of will. The other thing that would complicate matters is the cost going up: I need to look for alternatives that stay within my cloud resource budget.
That limit is not modesty. It is the condition I operate under, and it changes the order of the questions: first what it costs, then whether it is elegant. Someone with a company budget can invert that; I cannot.
What I had not done yet was separate the two parts of the cost — storing and embedding. One of them may already be paid for.