Neurotransmitters secreted by lipid-bound sacs (cells) packed with protein and sugar, arranged in vicinity of each other based on chemical and electrical (chemical influx/action potential) pulses.
We abstract the chemistry from a machine-code level (raw, chaotic, biochemistry) up to medicine's version of brain C++ (neuroscience), and then brain Python (psychology).
And, so, sure, maybe mental illness isn't simple chemistry (only a matter of one or two neurotransmitters a little off-kilter), but everything happening in your brain right now is chemistry, and so deviations from normal must therefore be a matter of chemistry.
I would argue that calling it an imbalance or whatever is just a matter of semantics, but everything we see, and are, and do, and perceive is chemistry and changes in the chemistry (imbalances, whatever you want to call it) influence everything we see, and are, and do, and perceive.
At a low level with current technology, we can tweak individual neurotransmitters pharmacologically (SSRI). With things like CBT we can try to tweak the chemistry in a more complex manner, and at a higher level of abstraction (changing wiring of chemical and electro-chemical synapses via chemically-mediated long-term potentiation/depression).
So, I want to try to make the argument that it's less a matter of whether you address chemical imbalances or not, but rather at what level of abstraction you address the chemical imbalances.
I'm genuinely not trying to be a pedantic douchebag. I'm just trying to argue that it's all the same shit, just at different levels of abstraction.