Add geom_smooth to part of of boxplot

I would like to add geom_smooth of each treatment with data only from week 0 to week 8 timepoints while keeping the "before treatment" boxplots in pre and post timepoints.
I have the following dataset:

timepoint <- c(rep("pre", 5), rep("post", 5), rep(paste0("week", 0:8), each = 4))
timepoint<-c(rep(timepoint,4))
treatment <- c(rep("before treatment", 10),rep(c("A", "B", "C","D"), 9))
treatment<-c(rep(treatment,4))
setseed(1)
measurements <- c(runif(5, 1.3,1.5),runif(5, 0.2, 0.4),runif(5, 0.3, 0.5),
runif(5, 0.4, 0.8),runif(5, 0.5, 0.7),runif(5, 0.7, 1),
runif(5, 0.9, 1.2),runif(5, 1, 1.3),runif(6, 1.2, 1.4),runif(5, 1.3,1.5),runif(5, 0.2, 0.4),runif(5, 0.3, 0.5),
runif(5, 0.4, 0.8),runif(5, 0.5, 0.7),runif(5, 0.7, 1),
runif(5, 0.9, 1.2),runif(5, 1, 1.3),runif(6, 1.2, 1.4),runif(5, 1.3,1.5),runif(5, 0.2, 0.4),runif(5, 0.3, 0.5),
runif(5, 0.4, 0.8),runif(5, 0.5, 0.7),runif(5, 0.7, 1),
runif(5, 0.9, 1.2),runif(5, 1, 1.3),runif(6, 1.2, 1.4),runif(5, 1.3,1.5),runif(5, 0.2, 0.4),runif(5, 0.3, 0.5),
runif(5, 0.4, 0.8),runif(5, 0.5, 0.7),runif(5, 0.7, 1),
runif(5, 0.9, 1.2),runif(5, 1, 1.3),runif(6, 1.2, 1.4))
mydata<-data.frame(timepoint,treatment,measurements)
mydata$timepoint<-factor(mydata$timepoint, levels=c("pre","post","week0","week1","week2","week3","week4","week5","week6","week7","week8"))

and the plot

ggplot(mydata, aes(timepoint, measurements))+
geom_boxplot(aes(color=treatment, fill=treatment))

I tried to add geom_smooth to this plot, but I suspect this did not work because the x-axis was not numeric. I also tried to create another column of mydata$numeric.timepoint with "pre" and "post" as -2, -1 and others as numbers without "week", but I would like to keep the x-axis breaks as in the figure.

You can achieve your suggested layout like this:

library(dplyr)
library(ggplot2)
mydata %>%
ggplot(aes(timepoint, measurements)) +
geom_boxplot(aes(fill = treatment)) +
geom_smooth(data = . %>% filter(!timepoint %in% c('pre', 'post')),
mapping = aes(group = treatment, color = treatment), 
se = FALSE, show.legend = FALSE
)

which is quite overloaded:

To serve the chart's purpose (presumably: get an idea of variation, and compare trends among treatments), you could try such:

library(dplyr)
library(ggplot2)
library(forcats) ## for convenient factor releveling etc.
mydata %>%
## reorder factor timepoint so that 'post' comes last:
mutate(timepoint = fct_relevel(timepoint, c('pre', paste0('week', 0:8), 'post'))) %>%
ggplot(aes(timepoint, measurements)) +
## show datapoints, slightly jittered and coloured
geom_jitter(aes(col = treatment), width = .1, height = 0, size = 1, alpha = .5,
show.legend = FALSE
) +
## show one smoother - but only CI band - per treatment
geom_smooth(data = . %>% filter(!timepoint %in% c('pre', 'post')),
aes(timepoint, measurements, group = treatment, fill = treatment),
col = NA
) +
## show overall trend - but without CI band - to compare treatments against:
geom_smooth(data = . %>% filter(!timepoint %in% c('pre', 'post')),
aes(timepoint, measurements, group = 'all'),
col = 'white', linewidth = .5
)

sidenote: should 'pre' and 'post' values be labelled the other way round?

<strike>It won't be easily possible if you want to plot everything on one panel. You would need to calculate the smoothing lines and then plot the lines through them (probably using splines for smoother curves).</strike> _{[proven wrong by user I_O]}

Maybe you want to consider facets. As a bonus, this is also de-cluttering your graph.

And due to the discrete nature of your x variable, you would need to specify a group aesthetic of 1 (or: group = treatment)

ggplot(mydata, aes(timepoint, measurements))+
  geom_boxplot(aes(color=treatment, fill=treatment)) +
  geom_smooth(aes(group = 1)) +
  facet_grid(~treatment)

### comes with lots of warnings due to your data

<!-- -->