Create API endpoint for making modifications to game world

[Volte6]

Description

This is just something I'm mulling over at the moment.

The general idea is:

• Create an API endpoint for world modification.
  • Should probably be RESTful, or mostly RESTful.
• All world-or-data modifying admin commands would behave as they do now, except final "writes" would be directed through the API.
• Additional tools could be written to interact with this API.

This have the benefit of de-coupling the admin interface from the code - it could even be moved to its own repository.

This would also allow other creative implementations to ultimately supplant the default tools, such as:

• A native application.
• A pure CLI interface

Concerns

• Creates extra maintenance and effort for modifying or creating new in-game commands
• Probably still want to lock game-state while changes or requests are being processed through the API

[sitzmana]

I’ve prototyped a basic RESTful API for running admin commands. Right now each request still needs an in-game user and room context, which forces an actual admin to be logged in. I’m considering two alternatives:
1. Introduce a dedicated “API” user (never shown in-game, always online) that carries full admin rights.
2. Split out headless versions of our admin commands that don’t require a user context at all.

I’d love your thoughts on which makes more sense for GoMud’s architecture and security model.

you can view what I have so far on my fork

[Volte6]

Ah I see, you're having it run the admin commands directly as though they are in-game commands?

Sounds interesting! That definitely would help prevent duplicating logic.

How does it handle the prompt system, which interrupts phases of the command?

What about using the admin user (UserId=1)?

Alternatively, you could probably spoof a UserRecord{} instance that is passed just to the function being called..

[sitzmana]

The spoofed UserRecord seems to work thanks for the tip! When I was trying to get the admin user, the users.GetByID returned nil if they were not in game.

I'm unsure about the prompt system yet, perhaps if you know the prompt that would come you could preload responses in the json. i.e. {"command":"item","name":"Sword","damage_dice":"1d8","confirm":true} or something similar

alternatively the http response could provide the prompt and wait for you to POST a followup, which could even interact with a curl wrapper to make it more interactive. I have not gotten that far yet though

[Volte6]

I think pre-populating the responses somehow would work.
We may even be able to use an ordinal slice with the responses if we can rely on an accurate process order. This would require some changes to the prompt system, so initially, I think key/value pairs of Question=>Response would make more sense, and preload that into the UserRecord.activePrompt somehow.

It's a tricky problem, but it's probably doable. The nice thing is if it fails, it will be out of the socket connection loop so shouldn't be destructive causing a hung connection or anything.