This is the "first" of a few iterations of this idea I can show. The idea is:
You have a constant combinator with minimum values.
You have a constant combinator with maximum values.
You have a "memory".
You have an "inventory". An inventory has "all components of interest" in it.
You have a "crafting machine".
For each symbol in inventory:
add 1 to the symbol in memory if inventory < minimum value and the symbol is not in memory.
remove 1 from the symbol in memory if inventory > maximum value.
The crafting machine is aware of memory and has set recipe so it will build anything on it. What this means is that by loading the constant combinator with 400 inserters, 200 fast inserters, and a minimum value of 50 inserters, you can build that run without seeing machine latency or switching.
You'll notice I've been vague about inventory and crafting machine. In the MVP the inventory is a space platform hub. Crafting machine is an assembler 3. However, you can place this on a robot logistic network. I use a proxy assembler to manage ingredients (a disabled assembler with set recipe). Anything that can set a recipe is a crafting machine and there's certainly a use case for building generic modules.
Designs can be provided with additional interest.
With a bit of effort, the RS idea can be adapted to use a train & radar combo. I've already done an iteration that used a tank poorly enough I won't show it, but well enough I think you can make your life a little easier if there's a logistic embargo on. The thing about a tank is that once it has vehicle logistics it can request all components.
My favorite adaptation is to put a selector combinator on it so it reads symbols based off a numeric index-symbol pair I give, so that means I can do assemblers then chemical plants. All this requires setting up constant combinators to work, but compared to adding a factory of machines it is a lot more positive. My hat is off to the devs for setting up a build table where the MVP will just work.
Practical Implementation
Screenshot 2025-09-23 191556b.png (823.56 KiB) Viewed 188 times
This is suitable for building a large pool of parts where individual requests are below ~400 units. It can, in theory, do all components itself by doing priority. In practice, I just imagine all the gears for a run of 500 underground belts. And I will commit that many belts to ideas like that. In other places, material handling will enforce limits on you: if you want quality EM Plants, you can move the idea to the ground but 150 plates showing up at once probably means you don't want to do it on a generic assembler. A single EM plant that makes 50 tier 2 modules over a few hours is worth a lot if you have nothing else. But it simply can't do what dedicated work station can.
On a platform with advanced asteroid processing, it saves tremendous space and lets you do bulk inserters without requiring extra rocket launches.