Generate books based on AI behaviour

[scotchfield]

This project accepts a JSON world as input, picks an agent inside that world, then lets that agent walk around while telling a story as they go.

https://github.com/scotchfield/agent-book

A sample book is here:

https://raw.githubusercontent.com/scotchfield/agent-book/master/clue.txt

Here's some sample text from the "Clue" world.

Call me Col. Mustard.

I feel like I've been here before. This is The Billiard Room. I see the lead pipe is here with me. I decide to leave, and go south, taking me to the conservatory. I start to feel sleepless. This reminds me of a statement I once heard, and I ponder it.

A sudden noise from one of the nearby rooms surprises me, and I suddenly feel myself feeling quite unimaginative. I'm standing in The Conservatory. As I look around, I see the knife in the room with me. I start to feel prerequisite. I decide to leave, and go through the secret passage, taking me to the lounge. I start to feel self-taught. The room looks different than I remember. Somehow, it seems more suspenseful.

I feel like I've been here before. This is The Lounge. I see the gloves is here with me. The room looks different than I remember. Somehow, it seems more spineless. Rev. Green is here with me, looking rather religious. Eventually it's time for me to go, and I leave toward the hall. After thinking about my surroundings, I begin to feel alright.

There's probably a lot more that can be done with this! Agents should be able to pick up objects, should have more Sims-like internal emotions, and so on. But it's neat to populate a JSON file with information about the world (like a MUD, with rooms and objects inside those rooms), then to pick an AI inside the world and see what they see as they move about.

NaNoGenMo is great and awesome. :)