I have data that is organized into distinct groups such that each group has its own regression line, and I expect the lines to be related, such as described in the first paragraphs here:

I have prior information on group level. For some groups I have strong priors, for others I don’t.
In the cases I’ve seen so far prior information is encoded only into hyperparameters, but I want to be able to specify what I know about some of the individual groups as well. Is that possible?

I guess you could make the priors for the group coefficients be a combination of the prior for individual groups and the hierarchical prior

data {
int N;
vector[N] slopes_prior_mean; //set to 0 when you know nothing
vector<lower=0>[N] slopes_prior_certainty; //set to 1 when you know nothing
}
parameters {
vector[N] slopes;
real<lower =0> slopes_prior_sigma;
}
model {
slopes ~ normal(slopes_prior_mean, slopes_prior_sigma * slopes_prior_certainty);
}

The only problem I see with that is that slopes_prior_certainty is hard to interpret.