I have an intuitive idea that, given some set of formulas $Γ$, and two formulas $A, B \not\in Γ$, $((Γ\cup{A}) \models B)↔(Γ \models (A→B))$. I can rationalize this as, if the left side of the biconditional is true, then whenever $A$ is true, $B$ is true, and if $A$ is false, nothing about $B$'s truth value is said. The same thing is descriptive of the RHS, and so they evaluate to true if both do and not true (false) if the other does.
How would I formalize a proof of this property, or express it in a rigorous fashion (in the language of logic)?
I'm aware that the logical connectives have valuation functions, such that, given valuations for the propositional letters/variables, any formula can be evaluated. But to my understanding, logical entailment ($\models$) is a metalogical concept and is not modeled the same way, but I cannot see any reason why it should not be.
Should the property be proven by creating an evaluation function for $\models$ as a connective, and prove the property for all valuations of $Γ, A,$ and $B$?
I feel that any answer to this will allow extrapolation to understanding other such metalogical properties as well, so feel free to be more general with your answers, but any insight into this particular problem will be greatly appreciated as well!