In many books about conformal field theory, when we talk about a coset $\mathfrak{g}_k/\mathfrak{h}_{k'}$, we would talk about how the modules of $\mathfrak{g}_k$ are decomposed into those of $\mathfrak{h}_{k'}$ tensoring those of $\mathfrak{g}_k/\mathfrak{h}_{k'}$, for example, for the vacuum module $$ \mathcal{R}_\text{vac}[\mathfrak{g}_k] = \mathcal{R}_\text{vac}[\mathfrak{g}_k/\mathfrak{h}_{k'}] \otimes \mathcal{R}_\text{vac}[\mathfrak{h}_{k'}] \oplus ... $$
I wonder if there is an "extra procedure" where we actually single out the vacuum representation $\mathcal{R}_\text{vac}[\mathfrak{g}_k/\mathfrak{h}_{k'}]$ from the sum, namely, a procedure where we actually get the vertex operator algebra $\mathfrak{g}_k/\mathfrak{h}_{k'}$ alone? (maybe it's as simple as focusing on states annihilated by some $\mathfrak{h}_{k'}$ generators, however I'm not entirely sure)