You need to solve the material-cost problem first. Either a space gun / elevator or ISRU. Otherwise, assembling on the ground is cheaper for everything but the most massive objects. (At which point a constellation or fleet usually makes more sense.)
> would allow to use well tested rockets to launch smaller parts instead of trying to build a huge rocket at once
Bigger rockets have moderate economies of scale. Optimising for a particular launch vehicle doesn’t make sense in the long run.
The farther humans go into deep space, the more important it will be to generate products with local materials, a practice called in-situ resource utilization.
Shelby finally retiring has been a great boon for NASA's human space exploration programs.
Also, technically the Lunar Gateway would fit your second point, considering that it's split into a separate habitation module and power+propulsion element.
Manufacturing seems possible but is several orders of magnitude more difficult because many processes on earth simply aren’t viable without access to earths vast array of manufacturing infrastructure, organic chemicals, or even gravity.
I can imagine someone making rockets from multiple separately launched parts that are then assembled in space. It would allow to use well tested rockets to launch smaller parts instead of trying to build a huge rocket at once.