This is a repost of my previous question, where I messed up everything. I make a explicit question here.
Suppose $\Bbb Q(\alpha, \beta)$ is a finite extension of $\Bbb Q$, where minipoly of $\alpha$ over $\Bbb Q(\beta)$ and over $\Bbb Q$ are different. How can I calculate the galois group by directly considering conjugate elements and write out all automorphisms that send $\alpha,\beta$ to conjugates where possible? i.e. I want to give an explicit construction of such galois groups.
For concrete examples you can consider $\Bbb Q(\sqrt2, \zeta_8)$
Note that I know $\Bbb Q(\sqrt2, \zeta_8)=\Bbb Q(\zeta_8)$, but I believe this is a coincidence and I can come up with another example for now. I want some sort of general solution to this kind of problems by directly considering conjugate elements, and $\mathrm{Gal}(\Bbb Q(\zeta_n)/\Bbb Q)\cong(\Bbb Z_n)^*$ doesn't provide such a solution to general cases. Is it possible?