As mentioned here, I’d like to get the log-likelihoods of the response based on the posterior draws of the model parameters in order to get `loo`

working for greta models.

Say that we have the following simple lm:

```
intercept <- normal(0, 1)
slope <- normal(0, 1)
sd_resid <- cauchy(0, 1, truncation = c(0, Inf))
x <- runif(100)
y <- as_data(rnorm(100, 1 + 2 * x, 1))
linpred <- intercept + slope * x
distribution(y) <- normal(linpred, sd_resid)
```

Is there a way to grasp from the y object the following:

- the distribution
- all direct parents objects

The aim would be to get a function like @nick suggested:

```
loglik <- density(y, log = TRUE)
```

Given that I never worked with R6 classes before some pointers would be much appreciated