- Does it mean $y=y(x,y)$?
No, it doesn't. As pointed out we have for example the ODE $y'(x)=\cos(x)+y(x)$. That means on the right hand side we have some dependency on $x$ and $y(x)$. Therefore we write shortly $$y'=f(x,y)$$ where the first component indicates the variable of your ODE and in the second component we have the function $y$ that dependents on $x$. There are also ODEs of higher order, for example $y''(x)=y'(x)+y(x)+4x^2$. Then we write shortly $$y''=f(x,y,y')$$ where in the example we have $f(x,y,y')=y'+y+4x^2$. I think you get the idea. Generally for an $n$-th order ODE
$$y^{(n)}=f(x,y,y',y'',...,y^{(n-1)}).$$
- Isn't an ODE always a function of one variable?
Firstly an ODE is no function. It is a special type of equation I'd say. And as seen it also depends on the solution $y$ and its derivatives. But in some way you are right - the solution depends solely on one variable. If it depends on more variable we are in the setting of partial differential equations.
- For instance here they use $y'=f(t,y)$. How should I interpt it?
Don't be irritated by this. Now we just have $t$ as the variable. For example $$y'(t)=t^3+\sin(y(t)) \rightarrow y'=f(t,y)$$ where $f(t,y)=t^3+\sin(y)$. We can freely choose the variables, have a look at the differential equations
$$\begin{align} g'(h)=h^2+\ln(g(h)) &\rightarrow g'=f(h,g) \\ \theta''(\zeta)=e^{\zeta}+\theta(\zeta)+2\theta'(\zeta)&\rightarrow \theta''=f(\zeta,\theta,\theta') \\ \kappa'(\mathcal{O})=4\mathcal{O}+\kappa(\mathcal{O})^2 &\rightarrow \kappa'=f(\mathcal{O}, \kappa)\end{align}$$