Improvement
Patch notes · November 10
The complete notes for this update.
Improvements
- Following your feedback in https://discord.com/channels/903327676884979802/1428769506875932814, we’ve made considerable changes to the memory system!
- New Memories are now created by a new and improved fine-tuned Mistral NeMo made by our very own model cook, ! Alpha testing has shown that it has significantly more detail retention and accuracy than our previous model.
- New Memories are now created every 6 actions instead of 4. This should ensure that there are fewer irrelevant Memories and give the Memory model a bit more context, leading to fewer mix-ups.
- Memory length is now increased to ~75 words instead of the previous ~50 to match the increase of the memory action window.
- The AI Instructions used by the Memory model have been significantly improved. Thank you for helping us on the process of improving the Memory system, next up is Summary! ### Other Bug Fixes
- Fixed an issue where making a multi-choice scenario would never load.
- Improved worker performance through multi-threading and resolved race conditions.
- Fixed an issue where you could not published adventures.
- Fixed the “shiny reward spinner”.
- Many text input areas in the app will now adjust in height depending on how much text is in the box.
- Fixed an issue with plot component cursor jumping to the end. --- *As always, if you encounter any issues or have feedback about these changes, let us know by posting them in the* [bugs-feature-requests channel](https://discord.com/channels/903327676884979802/1065040361057177702) If you are interested in Alpha and Beta Patch notes or would like to help test improvements earlier, feel free to select the patch notes notification roles here: <id:customize>