Set any Thing as a trap and arm it by providing the required ingredient. Each trap independently has a 50% chance of triggering when an enemy attacks you while in the area.
Misc:
Minimum level requirement for main quest: This incentivizes the player to explore further. Can be turned off under options
Fixed issue where only str and dex were generated as item attributes
Customizable negative prompt for local stable diffusion and novel ai image gen
Speculative fix for local stable diffusion periodically not working, by simplifying and using an sd-dedicated python exe as opposed to a wrapper exe
Blocked Places
Blocked places: Each place has a chance of spawning a blocked/locked place which can be unblocked via items/abilities for XP. These places also have a level jump which serves as a means to get to a higher-level area more quickly.
Misc: Fixed stable diffusion not working Tentative fix and extra debug statements for an issue with local A1111 stable diffusion only affecting a subset of users
A1111 for local image gen, better local text gen support
This update is mainly geared towards people who generate images or text locally using their NVIDIA GPU.
Overhauled the local image gen to use A1111 and its API. It generates images much faster than the original stable diffusion, and with less VRAM required!
Oobabooga API was not suitable for my text gen purposes because of the overhead in traditional event checks, so instead of going with that, I fixed some code in gpt_server to autoload the models better, so more models should be supported than before.
Image gen and text gen are also a little more transparent than before with the python code exposed under the "ai" directory, and being run via an included standalone python exe. This may also make it slightly easier to debug issues.
I also implemented preliminary support for Llama-based chat models, but this has only been briefly tested on my own computer, and only with the model Aitrepreneur/wizardLM-7B-GPTQ-4bit-128g which is small enough to fit on my GPU.
Difficulty Overhaul, Auto-generated UI image, and other fixes
Difficulty overhaul
Conversation was OP, so modified that to be influenced by charisma and pre-picked skills
Critical failures are now disabled for conversation, use consumable on self, and offer item, but can still be enabled in options if you enjoy the hilarious results
Roll checks are no longer so punishing in the beginning nor so easy when you’re leveled. Instead of being absolute, all roll checks now scale with the target’s level as well as being influenced by the item/ability level if applicable.
Generated image for exploration area
In the UI the left panel now gets an auto-generated image, which can be toggled off in the options.
Misc
Added ability to outright delete summaries or key plot points Added ability to delete NPCs Fixed flashing texture issue Added UI clarity for the ability to continue story without input without needing to spend any essences Fixed display names not updating after editing (finally) Added option to reset summaryception settings to defaults Fixed text log frame modding Fixed issue where enemy disappeared if http error occurred during location change with enemy Added a config option to prevent local AI from loading at start (only for debugging for people who experience CTD or freezing on game start) Fixed Open Assistant equipment detection Hopefully fixed issue where 6b model calibration would fail Fixed open assistant empty required skills in some cases
Notice to modders: Please be aware that some prompts have been changed
Open Assistant fixes, save during combat, possible fix for memory leak
Fixed open assistant not working after a patch they made which changed the way chats should be retrieved
Added button to delete side quests
For some event checks which are chatGPT compaible, ChatGPT now respects event trigger sliders, only if it’s dragged all the way to the left to the minimum value (you can think of it as turning off the feature).
Mitigated infinite death loop issue by increasing survival bar value when dying
Slightly increased font size for journal and modal views
Mysterious memory leak / crashing issue: Identified and fixed the most likely cause of memory leaks and/or crashing. It is thought to be caused by a particular type of texture allocation. All the code is now migrated away to the new way of allocating textures. This results in a brief flash before the texture is loaded, which will probably be fixed in the next update.
Fixed issue where defeating all the enemies in a room while using Open Assistant caused an error when trying to regenerate enemy names
Mod logic: Mods are no longer enabled by default when a new mod is detected. It defaults to disabled.
Improved summarization prompt to reduce chance of ChatGPT saying “protagonist”. Special thanks to Furoggu.
Added ability to save while in combat.
Moddable UI, player-controlled lorebook, other fixes
Most of the images of the UI can now be replaced via mods, so please mod to your heart's content. Examples in default/ui/info.txt. Please note, the default directory only serves as an example and is not being used, because I realized that loading images from disk instead of baking it into the game increases the load time by a few seconds. Custom images will only be loaded if they're in a mod directory other than "default".
Customizable preambles for ChatGPT: The "preamble" messages are now available in json form under chatgpt-preambles. Jailbreaking isn't needed for most scenarios, but it's there in case you need it. Please try not to increase the token count too much (a reasonable limit might be implemented in the future)
Player controlled lorebook: An optional text field now appears in the story tab view. It serves as a permanent player-editable grounding point for story facts, lore, important events, etc.
Fixed some of the most egregious Open Assistant issues such as constantly being attacked by non-existent entity
Misc - Fixed bug where summary couldn't be turned off - Inventory sorting preferences is now remembered in the save file
Starting attributes/skills + Open Assistant integration (experimental)
Starting attributes/skills:
The new game window now allows you to specify starting attributes and skills. The starting skills have also been woven into the preset scenarios as well as the auto-generation of those scenarios. This allows you to better roleplay your character and not be so helpless in the beginning.
Open Assistant:
Open Assistant (not to be confused with OpenAI) is a new free service open to the public, which does not have an official API. There is now an option to use the API in an unofficial way (subject to breakage if they patch it). Please note, even though I have provided a default cookie, this cookie may expire in which case you need to follow instructions on how to register your own free account and copy/paste the cookie.
The quality is not as good as OpenAI's ChatGPT, but is mostly better than the other free cloud option (kobold horde). The service reliability depends heavily on the time of day. More updates will be coming in the near future which may mitigate the queue time issue using better retry logic.
Misc:
Fixed some bugs in resource tracking Fixed widescreen equipment issue Fixed blank ability/item issue Fixed use ability on self issue in non-chatgpt mode
Paving the way for free/cheaper chatGPT (not really working yet) + various fixes
Began implementation of free ChatGPT-like functionality for Kobold Horde. This uses the new config line KOBOLD_HORDE_CHATGPT with a default whitelisted substring of GPT4-X-ALPACA. Warning: Currently this feature is so slow to the point of being unusable because there’s only 1 chatGPT model in Kobold Horde with high queue times. Also, there is a possible issue with the kobold horde API causing bugged outputs. I will continue monitoring the feasibility of various free/cheap ChatGPT-like options.
Abilities cooldown logic: Abilities now use a cooldown feature alongside energy requirements, instead of requiring a heal to recharge.
Fixed bug where if AI declared enemy defeated during an interaction with an NPC, the NPC would die
Fixed bug where no XP was gained if “outcome wasn’t great”
Added feature allowing you to use an ability to satisfy “required item/thing” in quests
Track raw resources
Track raw resources:
Press the dog paw symbol when recipe-based crafting is active to start tracking all raw resources for that recipe. Open currently tracked recipes by pressing the dog paw during inventory view.
Only tracked resources will be eligible as overriding raw resources for foraging and looting. Also, the probability of acquiring a raw resource (if tracking it) has been buffed. Please note: tracked items still need to pass the eligibility check to show up for foraging/looting merchants. For non-chatGPT, this is via the "[item] can be found in:" words. For chatGPT, it asks ChatGPT whether it makes sense for an item to be found in the particular area.
Misc
Fixed use ability on self bug
Customizable save location
Fixed new quest location even when quest location was already reached
Tweak ChatGPT checks including quest completion via story, self injury, enemy injury, and more
Reduced ChatGPT plot armor from 2 to 1. ChatGPT has a tendency to wish to continue a story rather than end it.
Fixed Wombo issue. Wombo changed VFX v2 into a premium model. Switched it to realistic v2.
Fixes incl better popup UI
Implemented tooltip big image for hovering over player
Fixed modding order being overwritten by writing to player prefs instead. Hopefully this update would be the last time anyone encounters overwritten mod orderings.
Fixed some false positives in all modes of text gen by preventing the used item/ability from showing up as a new item/ability if their names are almost equivalent
Fixed UI issue where tooltip would still show when modal window popped up
Fixed survival bar bug where survival bars weren't leveling up correctly (this fix works on older saves but cannot retroactively apply level ups)
Improved auto-scroll logic such that it stops auto-scrolling if you clicked in the window or scrolled yourself
Overhauled modal window UI: Moved it up to not block log text; also, player can now interact with non-turn-taking buttons behind the modal including scroll through the text log.
Fixed some false positives with ChatGPT via a new prompting technique which confirms whether item/ability was actually legitimate
Fixed a few formatting issues with ChatGPT generated entities
Reduced chance of abstract items in required item/location in ChatGPT-generated quests
Fixed ChatGPT issue where some enemy loot were presenting as "I'm sorry / need more information" etc due to prompt being unclear
Fixed ChatGPT raw resource implementation
Fixed survival bar wrong texture
Added retry logic for OpenAI. When there's a failure, it will retry once. This brings the chance of a failure from about 1/100 to 1/10000
Maybe fixed issue where NovelAI image gen would randomly stop working
