It is an enzymatic reaction - and comes from fireflies, or similar biological effects. So there's no reason that - at some point - the effect cannot be entirely generated in vivo rather than upon addition of particular ingredients. The hard part is getting enough light. Enough light to read by is a lot of energy for otherwise very efficient biological processes. Once engineered though, as long as the plant is energetically productive (alive and well-fed), the glow should persist. The glow could persist as long as the plant is alive.
The paper:
If the duration of the effect is only hours and then it runs out like rave glow sticks, is this tech still in a fetal stage? To fulfill the vision, an even more radical and novel approach is necessary to enable a "lightbulb plant" to be self-sustaining and not burn out for years.