Generating predictions for Dirichlet-multinomial

Modeling
1

I am trying to generate predictions for a Dirichlet-multinomial distribution in Stan (for a saved custom model).

When I save this code, I get an error on the dirichlet_rng(alphas) because I am sending dirichlet_rng(vector[ ]) and not dirichlet_rng(vector). What should my code look like? Would it be better to do a simple loop across all posteriors?

// This is testing implementation of a Dirichlet-multinomial model
functions {
}
data{
    // Structural
    int<lower=1> n_obs; // Number of observations
    int<lower=2> n_cat; // Number of categories

    // Population
    vector[n_obs] x_new; // Input variable x 

    // Posteriors
    int<lower = 1> n_iter; // Number of iterations for my samples
    
    // Population effects parameters
    real b_group_1_intercept[n_iter];
    real b_group_1_x[n_iter];

    real b_group_2_intercept[n_iter];
    real b_group_2_x[n_iter];

    real b_group_3_intercept[n_iter];
    real b_group_3_x[n_iter];

}

transformed data{
    vector[n_cat] alphas[n_iter * n_obs] ;
    //vector[n_cat] responses[n_iter * n_obs] ;
    for(iter in 1:n_iter)
    {
        for(obs in 1:n_obs)
        {
             // Linear Predictors for each group
            real linear_predictor_group_1 =  b_group_1_intercept[iter] + b_group_1_x[iter] * x_new[obs];
            real linear_predictor_group_2 =  b_group_2_intercept[iter] + b_group_2_x[iter] * x_new[obs];
            real linear_predictor_group_3 =  b_group_3_intercept[iter] + b_group_3_x[iter] * x_new[obs];
            
            // Apply the link function to each group's linear predictor
            real alpha_group_1 = exp(linear_predictor_group_1);
            real alpha_group_2 = exp(linear_predictor_group_2);
            real alpha_group_3 = exp(linear_predictor_group_3);
        
            // Turn alphas into a vector
            alphas[(obs-1) * iter + obs] = transpose([alpha_group_1, alpha_group_2, alpha_group_3]);

        }

    }    
}
generated quantities{
    vector[n_cat] responses[n_obs*n_iter]; // Response array
 
    // Get final part
    responses = multinomial_rng(dirichlet_rng(alphas);

}