I couldn't find the solution to the exercise for deallocation, so I wanted to share my own. Couple realization I had:

• there are typos in certain places, instead of returning batchref, it returns batch id. They're fixed in the code.
• sqlalchemy version is not pinned, so it'll try to install the latest one which is not compatible with the source code.(updated the requirements.txt)
• sku number is really not used anywhere logically for the deallocation method, as we cannot allocate anything to a batch if orderline and batch id doesn't match. we also know that orderline needs to unique, so data modeling doesn't really give any room to use sku for deallocation, I was not sure why it was included in the exercises to be used, maybe an oversight? I still used it as a placeholder but has no functional importance