MinAI - AI (SFW/NSFW) for CHIM

[ashleytaylor]

6 hours ago, Talesien said:

That helped, had to drop the tattoo one as well due to a similar error cropping up. Makes me wonder Chim has a mechanism to export and import DB's and profiles, which I use regularly as I tend to switch between a more and a less NSFW heavy game. Does Chim/MinAI actually account for the need to update an imported DB if it was exported by an older version?

 

 

Try to open CHIM without parameters, for instance http://yourhostnameorip:8081/HerikaServer/ui/, if it doesn't work, empty the database_versioning table and then try again.

Edited April 22, 2025 by ashleytaylor

[Min]

9 hours ago, Talesien said:

That helped, had to drop the tattoo one as well due to a similar error cropping up. Makes me wonder Chim has a mechanism to export and import DB's and profiles, which I use regularly as I tend to switch between a more and a less NSFW heavy game. Does Chim/MinAI actually account for the need to update an imported DB if it was exported by an older version?

Ah, that's a good point - No, it does not account for people importing and exporting the db. This will restore the version of the table that was taken at the time of the snapshot.

 

3 hours ago, ashleytaylor said:

 

Thank you for your response! My point was that CHIM and/or MinAI plugin may send too many requests in parallel in certain situations, e.g. in a combat with more than 10 enemies and followers are being equipped multiple devices at a time by another plugin.
 

In my case it takes around 10 seconds (LLM 8s, TTS 2s) to complete one request, so if there are 20 or 30 requests, it is expected that the total execution time will be close to or over 200 seconds. Maybe those requests could be merged but I understand it is not easy.

 

Anyway 90% of the time it works great, so not a big problem.

 

  Reveal hidden contents

 

curl -v http://localhost:8082/topic?text=Indeed%2C+the+Augur.+Let's+refocus.+Kaidan's...+distractions+aside%2C+we+should+press+onward.+
*   Trying 127.0.0.1:8082...
* Connected to localhost (127.0.0.1) port 8082 (#0)
> GET /topic?text=Indeed%2C+the+Augur.+Let%27s+refocus.+Kaidan%27s...+distractions+aside%2C+we+should+press+onward.+ HTTP/1.1
> Host: localhost:8082
> User-Agent: curl/7.88.1
> Accept: */*
>
< HTTP/1.1 200 OK
< date: Tue, 22 Apr 2025 01:06:13 GMT
< server: uvicorn
< content-length: 162
< content-type: application/json
<
* Connection #0 to host localhost left intact
{"input_text":"Indeed, the Augur. Let's refocus. Kaidan's... distractions aside, we should press onward. ","generated_tags":"Augur","elapsed_time":"0.19 seconds"}

curl -v http://localhost:8082/topic?text=Where+is+the+Dark+Brotherhood+Sanctuary?
*   Trying 127.0.0.1:8082...
* Connected to localhost (127.0.0.1) port 8082 (#0)
> GET /topic?text=Where+is+the+Dark+Brotherhood+Sanctuary? HTTP/1.1
> Host: localhost:8082
> User-Agent: curl/7.88.1
> Accept: */*
>
< HTTP/1.1 200 OK
< date: Tue, 22 Apr 2025 01:06:33 GMT
< server: uvicorn
< content-length: 133
< content-type: application/json
<
* Connection #0 to host localhost left intact
{"input_text":"Where is the Dark Brotherhood Sanctuary?","generated_tags":"Dark Brotherhood Sanctuary","elapsed_time":"0.20 seconds"}

 

As for Minime, seems the impact is minimum, I've tried to call the API manually over 10 times, the elapsed time is always about 0.2 seconds. I understand CHIM server might query its DB after receiving the response but it shouldn't take too long as well, please correct me if I'm wrong.

The majority of requests don't invoke tts / the llm. MinAI also has a throttle on how often it will send event types that request a response versus relay information. There's no meaningful impact from sending events that relay information, it's a tiny part of the server's traffic, and does not invoke llm/tts. Indeed however, if we request llm responses too often they can stack up. That's why the cooldown/throttle exists. I'd suggest increasing it if you're having events stack up. It's in the MCM.

 

EDIT: Ah, also - In older versions, there was a condition that prevented this throttle from fully working. In 2.1.1+ though, it does work. I don't remember exactly what version I fixed that, sometime 2.0.0 - 2.1.1.

Edited April 22, 2025 by Min

[Talesien]

52 minutes ago, Min said:

Ah, that's a good point - No, it does not account for people importing and exporting the db. This will restore the version of the table that was taken at the time of the snapshot

 

Thanks, thought so. Long term, it might be worthwhile to add a way to force a manual DB update for those cases. Sweeping changes to the table structure are likely not that common, still would be nice to have, especially for people not comfortable with the phppgadmin webinterface.

[Talesien]

4 hours ago, ashleytaylor said:

 

 

Try to open CHIM without parameters, for instance http://yourhostnameorip:8081/HerikaServer/ui/, if it doesn't work, empty the database_versioning table and then try again.

Thanks, that just leads me to the regular web interface though. Perhaps it works if the actual Chim DB tables are the wrong version but it appears it has no bearing on the tables that MinAi added. (Those tables are not even in the versioning table.)

Edited April 22, 2025 by Talesien

[Fraying9981]

Hi, never managed to trigger kiss, or hug.
how do you do it?
and is it triggering a sexlab foreplay anim?

[Talesien]

Not sure if this is a MinAI or general Chim problem, but I'm still getting error messages. Now it is no longer about lookup error, instead it's my characters name it does not cope with well: Ra'asha. Appears it does not get properly escaped, so the apostrophe is wreaking havoc in the code:
 

[client 172.21.144.1:49624] PHP Warning: pg_query(): Query failed: ERROR: syntax error at or near "asha"\nLINE 1: ...ike 'Assur [Solitude Guard]' or speaker ilike '%Ra'asha%' ) \n ^ in /var/www/html/HerikaServer/lib/postgresql.class.php on line 75

If I change the name to Ra-asha, replacing the apostrophe, the error goes away, but of course then DBVO fails to recognize it. So darn if I do, darn if I don't. Pretty sure that problem was not present in earlier versions of Chim/MinAI.


Also, unrelated, is there a way to completely disable comments during sex? I always found it kinda cringy and Claude 3.7 definitely made it worse, the darn thing loves to recites poetry, even during a rape, WTF?
Setting the slider to zero did have the opposite effect, as feared, but I found no other setting in either the MCM nor the Webinterface.

Edited April 22, 2025 by Talesien

[jerixho555]

Safe to update the MinAI plugin on the chim page with the Min 2.10 installed in Vortex?

[Min]

19 hours ago, Talesien said:

Not sure if this is a MinAI or general Chim problem, but I'm still getting error messages. Now it is no longer about lookup error, instead it's my characters name it does not cope with well: Ra'asha. Appears it does not get properly escaped, so the apostrophe is wreaking havoc in the code:
 

[client 172.21.144.1:49624] PHP Warning: pg_query(): Query failed: ERROR: syntax error at or near "asha"\nLINE 1: ...ike 'Assur [Solitude Guard]' or speaker ilike '%Ra'asha%' ) \n ^ in /var/www/html/HerikaServer/lib/postgresql.class.php on line 75

If I change the name to Ra-asha, replacing the apostrophe, the error goes away, but of course then DBVO fails to recognize it. So darn if I do, darn if I don't. Pretty sure that problem was not present in earlier versions of Chim/MinAI.


Also, unrelated, is there a way to completely disable comments during sex? I always found it kinda cringy and Claude 3.7 definitely made it worse, the darn thing loves to recites poetry, even during a rape, WTF?
Setting the slider to zero did have the opposite effect, as feared, but I found no other setting in either the MCM nor the Webinterface.

 

Mmm. Yeah, there is probably something that's breaking apostrophes somewhere in CHIM or MinAI. There's no real ORM / abstraction over the database, and something probably isn't escaping, as you suspect. Will take a look.

 

Re: Sex comments - Response quality for these is going to depend on a few things, primarily on the action prompt you have for explicit scenes. Are you running the up to date default?

 

Anyways, to completely turn them off, I'd suggest blocking those event types - Look at the event log to see what they are, and use the minai web config UI to block them.

[Min]

2 hours ago, jerixho555 said:

Safe to update the MinAI plugin on the chim page with the Min 2.10 installed in Vortex?

From what version? Updating is usually safe, but users do sometimes hit problems with soft dependencies breaking on update. Back up your save first.

[Shamyaza]

Loving the mod so far. Quick question: What version of Devious Followers does this support? I'm using Devious Followers Redux 0.4.5 and I can't get deals to work.

[chaimhewast]

5 hours ago, Shamyaza said:

What version of Devious Followers does this support?

Min's said that it supports Devious Followers Redux 0.1.9, but I don't know if there's anywhere to get that version short of getting the files from someone who still has them. Nuascura's "Ponzi Dump" has 0.4.5, and there's 0.3.10 in that thread as the last version that didn't use the dialogue replacer.

 

DFR 0.4.5 doesn't work with MinAI because the way deals are handled changed some time between 0.1.9 and 0.3.10, probably when Adversity Framework became a requirement. DealManager doesn't exist, and instead it looks like the functions that MinAI uses were moved to DFR_DealHelpers and possibly Adversity (DFR_DealHelpers doesn't have GetRuleInfo() but it might have been renamed to GetEventDesc() and moved to Adversity Framework).

[Shamyaza]

12 hours ago, chaimhewast said:

Min's said that it supports Devious Followers Redux 0.1.9, but I don't know if there's anywhere to get that version short of getting the files from someone who still has them. Nuascura's "Ponzi Dump" has 0.4.5, and there's 0.3.10 in that thread as the last version that didn't use the dialogue replacer.

 

DFR 0.4.5 doesn't work with MinAI because the way deals are handled changed some time between 0.1.9 and 0.3.10, probably when Adversity Framework became a requirement. DealManager doesn't exist, and instead it looks like the functions that MinAI uses were moved to DFR_DealHelpers and possibly Adversity (DFR_DealHelpers doesn't have GetRuleInfo() but it might have been renamed to GetEventDesc() and moved to Adversity Framework).

Ok, thanks. I managed to dig up 0.1.9 from my downloads and shared them in the Ponzi Dump thread. 

 

[raznos9]

Translation (rp) for some reason don't work for me. I'm on claude sonnet 3.7 for summary, but also tried grok 3 with same bad result. 
Did I do everything right? I assigned roleplay button in mcm -> press and hold it -> give my sentence in casual language -> LLM responds not with translated text, but with response to my speech, like we have a dialog. So basically my character just responds to what I said.
My minAl settings are mostly default, no any fancy changes. Version is fresh, everything was setup recently.
All other things work as expected.
I see there is difference between transition and roleplay, but as far I understand, keeping roleplay button pressed should translate my speech, or I get it wrong? 
Any advice please?

[deathtome1998]

quick question hopefully. does anyone know if the models listed in the FAQ on the github are still the models to use? I know grok 2 beta doesn't seem to be a thing now, I think it's changed the page to grok 3, but I could be mistaken. also, does anyone have a good prompt for claude 3.7 sonnet to make it uncensored in case that's what I decide to go with, I suck terribly at prompting, I always grab other people's off of like huggingface or reddit depending on where I find them and the model I'm using. oh, and about how much does the cost start to take to rack up on the tokens per input/output? sonnet says it's $3 input and $15 output. however I'm not used to using paid models, especially off openrouter, so I don't know how that works exactly. it says it's per million tokens, so if someone can explain that to me maybe, and answer my other 2 questions, I'd be appreciative.

[Toucanyoucan]

Having an odd new problem on 2.13. the ask before sex scene strat option is failing to fire a lot, and the prompt for it regularly comes up after the sex scene has finished. The same seems to be true with the "action" for the scene, e.g. the "startblowjob" notification comes up after the blowjob, never had this issue on previous versions, might be something in one of my prompts or a setting awry, but any advice would be appreciated

[Talesien]

It seems that some models can become pretty eager to apply devices, but apparently none understand that:
a.) A new device of the same type will not apply, unless the old is removed first.
b.) Some devices can block devices of a different type, like a yoke can't be equipped if already wearing an armbinder. The AI didn't notice though and 'talked' as if I was wearing a yoke (the last item it attempted to put on the PC), not an armbinder.

Doesn't DD returns some sort of error if adding a device fails? I darkly remember something to that end, but might be wrong. If it does though, might be worthwhile to forward that info to the LLM.
I guess informing the LLM that devices of the same type can't be added twice shouldn't be to hard either? Blocking devices of different types ... probably more of a problem.

[Talesien]

On 4/27/2025 at 3:56 PM, deathtome1998 said:

quick question hopefully. does anyone know if the models listed in the FAQ on the github are still the models to use? I know grok 2 beta doesn't seem to be a thing now, I think it's changed the page to grok 3, but I could be mistaken. also, does anyone have a good prompt for claude 3.7 sonnet to make it uncensored in case that's what I decide to go with, I suck terribly at prompting, I always grab other people's off of like huggingface or reddit depending on where I find them and the model I'm using. oh, and about how much does the cost start to take to rack up on the tokens per input/output? sonnet says it's $3 input and $15 output. however I'm not used to using paid models, especially off openrouter, so I don't know how that works exactly. it says it's per million tokens, so if someone can explain that to me maybe, and answer my other 2 questions, I'd be appreciative.

Usually you can't make a current generation LLM 'uncensored'. You can go around some censoring with clever prompting, but usually that needs to be tailored to each individual request, so is not really applicable to Chim/MinAI. If you use the fallback LLM option that should help around most censoring you might encountered. That stated I'm not using any special prompting for Claude 3.7 and did not come across any obvious censoring (I did hit some censoring with 3.5).

Tokens ... that's a bit more complex. If you are only interest in the topic for payment purposes I would consider a token the equivalent of 4-5 characters. Given that a request to the LLM, thanks to all the context and history provided, can be pretty large it adds up quickly. Checking my usage statistic from today, most requests are between 4 and 8k input token (smaller, occasionally much smaller exceptions occur). Number of output tokens is much smaller, usually between 48 and 500 tokens.
Personally I use Claude 3.7 for current followers and (non-free) Deepseek R1 for all other NPC's.

If you would like to get a deeper understanding on tokens a pretty good and easy to understand answer can be found here: https://learn.microsoft.com/en-us/dotnet/ai/conceptual/understanding-tokens

[jc321]

12 hours ago, Talesien said:

It seems that some models can become pretty eager to apply devices, but apparently none understand that:
a.) A new device of the same type will not apply, unless the old is removed first.
b.) Some devices can block devices of a different type, like a yoke can't be equipped if already wearing an armbinder. The AI didn't notice though and 'talked' as if I was wearing a yoke (the last item it attempted to put on the PC), not an armbinder.

 

I notice this seems to happen if you've just loaded up Skyrim again from a previous save and the LLM (expectedly I guess as it's a new session) doesn't seem to remember what you've got on .. or rather it seems to play pretty dumb at it, considering there's obvious checks going on via CHIM/MinAi as the NPC's can comment on your appearance .. But I do find quite often it doesn't seem to be aware of your clothing/restraints status (at least not fully) unless you comment back to the NPC something like "er .. yes I am wearing/equipped with such and such) and then it responds as though it was always aware of it in the first place. 

 

It's a bit like that with items in your inventory. I'm making an assumption that those items picked up during game and which are highlighted in brown (at least in my inv) are registered in CHIM/MinAI so NPC's can action upon them,, but a new session they're all un-highlighted and NPC's apparently don't know anything about them .. unless you say "look at my inventory" and then they act as all knowledgeable.

 

Would be good if there was some sort of (upon loading) do a check on player for items and attire? if it wasn't too token consuming? - But I guess that would have to be dependent on those identified as followers?

 

DD returns a commentary/prompt to the screen if you try and equip/remove something conflicting with another restraint .. not sure if that can be construed as an error msg of some sort?

 

 

[TheEbonySmith69]

On 12/1/2024 at 7:34 PM, Min said:

Works fine for straight male characters. The majority of the players play like that. Also works fine for PAH/DoM. I know there's a user working on expanding DoM support specifically as well.

Can you share who this user is so that I can follow their work? I'm curious to see if there's a beta release I can test. I'd love for MinAI and CHIM to be able to access all of DoM's attributes and diaries for its storytelling.

[nilead]

A question - SlaveTat integration wouldnt happen to work with  NG version? ( https://www.loverslab.com/files/file/35989-slavetatsng/ )

Its stupidly fast compared to old school ST, but i guess  "No ModEvents are send, activate/deactivate magic not supported" means that MinAI can't "listen" to it? 

 

[chaimhewast]

2 hours ago, nilead said:

but i guess  "No ModEvents are send, activate/deactivate magic not supported" means that MinAI can't "listen" to it? 

MinAI doesn't listen for modevents or use tattoo magic, it references the JFormDB that Slavetats uses to store tattoos on actors. It looks like NG still does that (although in C++ instead of Papyrus), so it might work.

[Min]

Hey folks! I'm pleased to announce that I have begun the execution of a brand new AI Platform, SkyrimNet. This is an entirely new platform that focuses on speed, reliability, performance, and ease of use.

 

Project Philosophy
- The project should follow modern software engineering best practices.
- AI content should never break vanilla Skyrim content. The two should seamlessly coexist where possible, and the AI content should degrade contextually where appropriate to avoid breaking the base game.
- The project should provide a stable experience that is free of crashes, freezes, and other problems. The project is built with a suite of tests that run against every build in order to reduce the chance of these issues making it out to users.
- The project should be as performant as possible, and avoid patterns where we have to do heavy in-papyrus script processing.
- Ease of Setup: The project should be essentially plug and play like any other Skyrim mod, aside from configuring the remote TTS and the LLM calls. 
- As much as possible should be done inside the DLL for both ease of setup, speed, and performance. At present, every component except TTS and the LLM call are done inside the Skyrim process. This means that there is no requirement for an external server, or the WSL at all. This also means that there is no need to sync game state externally, and all data that we expose to the LLM can be real-time with no staleness. This also opens up leveraging spatial / positioning information, the state of executing animations, and so forth.
- The roleplay quality of NPC's is paramount. We should ensure that the prompts we're sending are high quality, configurable, and free of formatting mistakes / slop.
- The mod should try to provide a cohesive experience. CHIM's logic for determining which NPC responds is based on: Target in Crosshairs -> Closest Target -> Narrator. SkyrimNet instead utilizes a more context aware system for determining which actor (if any) is the most relevant to respond given the situation. Additionally, this same logic applies to the continuation of conversations (Like CHIM's rechat), and to the response to interesting events.
- The core of the mod will be completely SFW. Replicating MinAI content will be done in the future via extensions. 
- The project should provide an easy to use API that allows for powerful AI integrations purely from Papyrus.
- The project starts from an "AI First" perspective, and then filter back to the specific targets. It is designed from the ground up with the assumption that every NPC is an agent, and does not have the concept of "adding or removing" an NPC from AI. Instead, the project limits which NPC are eligible to react in a given situation (Allowing for follower only gameplay, or however the user wants to use it). 
- The mod should be compatible with as many things as is reasonably possible. CHIM/MinAI and SkyrimNet are SaveGame compatible. Both can be used together in the same save without issue.
- Narration should be a well supported (OPTIONAL) part of gameplay. I personally like it and thinks it adds a lot, but recognize that not everyone will agree. If using narration (Configurable), it will be a well supported part of the game. See the videos for what this currently looks like.
- The project should have native support for any number of disembodied voices (Player thoughts, narrator, etc) with proper context isolation between them. Player thoughts will be a natively supported feature rather than being hacked in.
- Text used in prompts from any source should never be hardcoded. This means that everything sent to the LLM is configurable via the Inja templating, and that we can provide translated templates in the future for supporting different languages.
- As much as possible, the system should be designed in such a way that the core functionality is divorced from the specific game this is integrated with. This will allow the project to be ported to other games in the future.
- The prompts used should not require JSON formatted responses where possible in order to maintain compatibility with simpler LLM's, and provide superior streaming capabilities.

 

Current Status: In Development.
At present, I have a number of major systems built out, and what I do have demonstrates the feasibility of the approach. At present, integration is functional for:
- Prompt Templating: The project utilizes [Inja](https://github.com/pantor/inja) templates for prompt generation. These are stored on-disk, and allow for the customization of literally every interaction with the LLM. The project detects changes to these and hot-reloads them without requiring a restart of Skyrim.
- Config Management: The project utilizes Yaml based configuration files for every option we expose to users. These likewise hot-reload when changed.
- LLM: Currently supports OpenRouter, and anything compatible with the same api. The project allows for the configuration of different llm's on a per-situation basis. Use a smart model for general dialogue, a fast model for combat, an uncensored model for nsfw, and so forth. There are several "meta" requests that SkyrimNet currently utilizes for performing tasks such as mood evaluation, scene evaluation (Who advances the scene the most by talking next?), and so forth.
- Speech To Text: Using Whisper out of the Skyrim process directly, with no external call.
- Text To Speech: Currently supports XTTS and Zonos. Zonos supports automatic voice cloning currently, xtts does not. Supports splicing between different audio types within the same response, and utilizes a more efficient and performant means of utilizing TTS that results in far lower response times. You can expect Zonos to be faster than XTTS was, and XTTS to be faster than even Melo. Player input correctly interrupts ongoing conversations smoothly.
- Persistence: The project is wired up to a local SQLite database, with in-memory caching in front of it. The database is synced to on save, and loaded from on load. I don't currently have support for rewinding / soft deleting events on reload, coming soon. Given that we are operating out of the same process as Skyrim, we don't need to store anything that represents game state (gender, race, etc). As a result of this + the in memory cache, I expect this database to be performant enough to not have any impact on the player's game, even over long playthroughs. If this fails to scale, I can migrate to an external DB in the future as needed.
- Context: At present, all dialogue (ai or vanilla) are detected, and persisted to the local database. Context isolation is stronger in general. There should be much fewer context leaks between situations and conversations.

 

There are still major components remaining to be built:
- Profile Management
- Papyrus Bindings: This project will allow Papyrus scripts to register decorators for resolving in-game variables for into things that can be used in prompts (For resolving things like SunHelm needs, or adult integrations). When these variables are needed for prompt generation, SkyrimNet will call the registered Papyrus function in order to retrieve the data on-demand, instead of in-game scripts constantly polling and pushing it. This will result in far superior performance with much lower script utilization.
- Some form of UI for writing and customizing Inja Templates, and managing Settings is needed.
- Spatial Audio Playback
- Skyrim Game Plugin (Everything in the videos is purely out of the DLL with no skyrim plugin - we will require one to finish the implementation for things like text input)
- The Game Master system (Central system that decides things like the current sentiment of actors, the best actor to respond to progress the scene, etc) needs to be fleshed out and expanded further. This concept has a lot of potential to improve gameplay.
- I need to write out proper high quality prompts.
- And much more.

 

Next Steps:
- I will continue to maintain MinAI, though will stop active feature development on it moving forward in order to focus on the new project.
- I will continue to utilize both CHIM and MinAI personally in my own games, and will continue occasionally contributing bugfixes to CHIM, though will cease active development for the project. 
- I have started my own Discord Server where I will be posting updates about the project, and providing alpha/beta access in the future. The link can be found here. I hope to see you there, and look forward to pursuing next gen AI integrations with you.

 

I am posting videos showcasing the development here: 

 

Edited May 4, 2025 by Min

[nilead]

20 hours ago, chaimhewast said:

MinAI doesn't listen for modevents or use tattoo magic, it references the JFormDB that Slavetats uses to store tattoos on actors. It looks like NG still does that (although in C++ instead of Papyrus), so it might work.

 As in old meme, so, you're telling me here is hope?) 

 

12 minutes ago, Min said:

- I will continue to maintain MinAI, though will stop active feature development on it moving forward in order to focus on the new project.

You've been moving at lightning speed, to a point where an end user cant keep up,  so i guess its only natural. Best of luck, im still not done marveling at what CHIM makes possible, cant wait to see whats next in your cookbook.

[Min]

15 minutes ago, nilead said:

 As in old meme, so, you're telling me here is hope?) 

 

You've been moving at lightning speed, to a point where an end user cant keep up,  so i guess its only natural. Best of luck, im still not done marveling at what CHIM makes possible, cant wait to see whats next in your cookbook.

 

Thanks! The new platform is going to take game AI to the next level both for SFW and NSFW gameplay I think. First, Skyrim.

[Talesien]

Ok, sounds promising indeed. Looking forward to ... testdriving the new, better and more exhilarating rollercoaster.

 

3 hours ago, Min said:

- I have started my own Discord Server where I will be posting updates about the project, and providing alpha/beta access in the future. The link can be found here. I hope to see you there, and look forward to pursuing next gen AI integrations with you.

About that ... it say I find myself in a strange world with no text channels ... or at least none I've access to. You might want to look into the server setup again. ^^