If you go back almost 50 years to the early Space Manufacturing Conference papers (they are online but expensive from the AIAA), and read the papers from Tom Heppenheimer, regolith was to be launched off the moon to a location near L2. Actually somewhat inside of L2 to compensate for the impact of packages of regolith. Tom described the impacts at the "catcher" as being like a stream of 18th-century cannon balls. When the catcher got full, it moved over to L5 using a rotary pellet launcher, throwing away some of the regolith for reaction mass. The mass driver location on the moon was very specific to take advantage of an "achromatic" orbital path, where small variations in the mass driver exit velocity still got the payload to the target. The mass driver switching in those days was SCRs. They had a fair amount of jitter in switching time and, as a result, variation in the exit velocity. Modern insulated gate power transistors are faster and more consistent.
You might be able to find those volumes in a library. I had two of the first 3 and found the missing one for sale, perhaps on eBay,.
Launching finished parts off the moon is an alternative, but the acceleration is brutal for a reasonable length mass driver. You also have to figure out a way to catch the parts or include rocket engines and guidance in each one.
An alternative that was not available in those days is a moving cable lunar elevator out through L!. It has to hang down in Earth;s gravity far beyond L!. If you make it 190,000 km long, dropping payloads off the end puts them in a Hohmann transfer orbit to GEO.
It is a complex set of tradeoffs.
Keith