I have a recent discussion about this problem.
Let me assume you already have the expectation value of the equal time anti-commutator
\begin{equation}
\langle \{ \psi( x ) , \psi^{\dagger}( x' ) \} \rangle = \langle :e^{i \phi(x) }:, :e^{ -i \phi(x' )} :\rangle = i \delta( x - x' )
\end{equation}
This can be established for example through analytically continuing the Euclidean boson correlator $\ln | z_1 - z_2|$ with $t = i \epsilon$ trick, or operator formalism.
For a single oscillator we have
\begin{equation}
:e^A: :e^B: = :e^{A+B}: e^{\langle A B \rangle }
\end{equation}
generalizing this to the vertex operator
\begin{equation}
\begin{aligned}
\{:e^{i \phi(x) }:, :e^{ -i \phi(x' )} :\} &= (e^{\langle \phi(x) \phi(x') \rangle } + e^{\langle \phi(x') \phi(x) \rangle }):e^{i (\phi(x) - \phi( x' )) }: \\
&= \langle :e^{i \phi(x) }:, :e^{ -i \phi(x' )} :\rangle : e^{i (\phi(x) - \phi( x' )) }:\\
&= i \delta( x - x' ) e^{i (\phi(x) - \phi( x' )) }\\
&= i \delta( x - x' )
\end{aligned}
\end{equation}
The last line is an unexpected way to reduce an operator to a c-number.