Word of warning!!!!! If you use LM Studio for Rimworld and Echo Colony your ram usage will go up. It will show Rimworld using a lot of ram. I suggest not loading a Model into LM Studio until your at Rimworld's Main Menu.

https://imgur.com/a/z1yZ1Iv should explain everything.

Gerik Uylerk posted this in another thread: https://imgur.com/4J6N0TY for setup.

To get to EchoColony's mod options, you need to go to Options>find or search for EchoColony and input the info.

Edit1: increasing evaluation batch size can help the ai remember past actions better. I know it works awesome in lm studio chat itself. I'll post an image sometime in the near future.

Update: 6/22
A higher Evaluation Batch Size may make LM studio run slower in Rimworld. There have been updates in some of the models.

tensorblock/neuraldaredevil-8b-abliterated
dirty-muse-writer-v01 by mradermacher are pretty good. These are the updated ones within the last 2 months.

AI Chat

The above two models work fairly well. Deeptalk takes too long to respond and Minstrel types very slowly IMHO.

When you get to around 10k tokens you should watch out for inconsistencies in your stories ect.... If this gets anoying turn on RoPE Frequency base and put it to around 12,000 and RoPE Frequency Scale between .5 - 2.0. A scale of 1 seems to work fairly well. If you are writing a story with many changes it can confuse the model.