6 ms·

I grew up coding MUDs (in college) and every year I try to recreate that glory, so I start a fresh MUD code base from scratch. This year, I am creating a nodej

I grew up coding MUDs (in college) and every year I try to recreate that glory, so I start a fresh MUD code base from scratch. This year, I am creating a nodejs MUD, well closer to a MUSH actually. * Everything is an Entity with Attributes * A player is an Entity with its client property set * There's a handful of built-in commands, but the important ones are @eval and @set-attribute * The soft-code language is ... JavaScript. * It uses the vm module from nodejs[0] to execute stuff. * It has separate clients (so far) for Discord, Slack, and local console. It's easy to add more and will be easier after I do this one refactor. Each time an soft-code command or function or what have you executes, it creates a new context to isolate the soft-code from the rest of the application and uses Proxy objects to isolate the database Entity from the soft-code Entity. As an example, this is the look command (placed on the global registry object): if (!me.location) { me.send("You are nowhere."); } else { me.send("**" + me.location.name + "**"); const contents = me.location.contents; if (!contents?.length) { me.send("You see nothing here."); } else { me.emit("You see:\\n" + contents.map(thing => "* " + thing.name).join("\\n")); } } The Proxy manages access to Entities and Attributes (for example, Attributes can only normally be accessed by their owners). The MUSH outputs markdown for everything, so each client can format it however it likes. I know the security is not nearly close to perfect, but it's mostly a proof of concept that I will make public one day. I did do a few things, like disable eval and Function and Promise but there's probably plenty of ways to get around it.

imtringued 1mo ago
Have you thought about letting an LLM generate the item and object scripts from a description? The LLM probably shouldn't be involved in regular game play though.
- kageneko 1mo ago
  Yeah, but that's a different project than this one :D This is just about building the framework for now and then putting it loose into the world. I have a big refactor in mind to further decouple the client/server from the game engine, but it'll take some work. I did make a sample database (a JSON file that was imported into the database) for an earlier iteration that was almost entirely LLM generated. I provided it with description of the general area and what sorts of things could be found there, along with the climate and such. It did a really good job with descriptions and added little touches but it had trouble keeping the map in mind and so things that were located in the north could also be reached by going east. That could also be prompting on my part or just not giving it a strong enough framework.

imtringued 1mo ago

Have you thought about letting an LLM generate the item and object scripts from a description? The LLM probably shouldn't be involved in regular game play though.

kageneko 1mo ago
Yeah, but that's a different project than this one :D This is just about building the framework for now and then putting it loose into the world. I have a big refactor in mind to further decouple the client/server from the game engine, but it'll take some work. I did make a sample database (a JSON file that was imported into the database) for an earlier iteration that was almost entirely LLM generated. I provided it with description of the general area and what sorts of things could be found there, along with the climate and such. It did a really good job with descriptions and added little touches but it had trouble keeping the map in mind and so things that were located in the north could also be reached by going east. That could also be prompting on my part or just not giving it a strong enough framework.