Hi Greg,
You are correct. You will need to define your allocation matrix in a portfolio payout function (like equal-weight). This portfolio payout function takes a multi-colum (assets) xts of asset returns as input, and returns an xts (same shape as input) of allocations per asset. Constraint is that total weights of the portfolio (equivalent to summing all values of one row in your xts) is between 0 and 1. As per individual strategy payouts, you will need to publish this portfolio payout in Qlib.
As you may need some past values to do your calibration, for example if you are using rolling historical asset volatilities or rolling asset betas to the benchmark, you may want to use an equal weight allocation (no need of past values) until you have enough data points to compute your actual allocation logics.
Hope this helps.