Is it possible to use Herika with Google's Gemini API? Their free-tier is remarkably generous with a 1,000,000 token context window and access to Gemini 1.5 Pro.
I haven't tried running a detached AI server (free), but I would expect it would take heavy resource load and the quality and conversation lantency probably wouldn't be as good - especially with likely lower quality voice/speech model.
GPT is super cheap though. I didnt enable the visual aspect, but basically I have run it with quite heavy & lengthy sessions with Mantella running at the same time and its like less than 10 bucks a month for both running off the GPT 3.5.
After using Herika I now want every NPC in every game to have AI, that's how good this mod is! I've had no issues with it, and as for compatibility it's a standalone follower mod, what else do you need? I've used other mods to customise her look and what she wears, and everything works perfectly.
This is absolute nonsense, you either did something wrong in installation or you are basically just grieving.
I"m running with about 900 mods and Herika. Her avatar is completely compatible with flower girls/romance mods, amazing followers tweaks, dozens of NPC tweaks and modifications, and just about anything else you can possibly think of that would in any way interact with Herika herself.
It is a bit challenging and I haven't come up with a solution to get it working with Mantella the way "I" would like her to. It works, with radiant conversations and everything else, but its like a different persona in Mantella than the Herika we all know. Basically she has multiple personalities between the Mantella Herika and the "regular" Herika. I am hoping to find a way to set her up so that she can partake in radiant confversations with Mantella driven NPCs with the "real" Herika. Other than that, which may not even be possible yet, Herika is compatible with everything. I wanted to add this post incase anyone is considering putting her in the game, so a nonsense post doesn't scare them off.
PS Antother thing I would really like is to be able to set up multiple "herika" servers running simutaniously so that my other followers can have the same robust AI with their own personalities. Again, I don't think this is possible yet but I'm not sure.
I'm running a 5600X, 3070Ti, and 32GB DDR4 at 3600. I get 2 frames per second. I think I need a less intensive LLM to get out of potato mode.
*edit - 1 week later* I started with a 13B LLM in an older format, the person in the video I used to get some starting advice recommended using one in a .gguf format even though his links started with the older one. I found a 7B LLM in .gguf, and what a difference. It ran so much better, but not perfect. Pretty sure I would need some upgrades to smooth it out. It was mostly some frame skips if I came near an NPC that I could talk to, and Herika was starting to listen.
Running stable at 60 fps 4K resolution: - AMD Ryzen 9 5900X - RAM 64GB - video RTX 3060 12GB - ENBoost to extend video memory used by Skyrim - KoboldCpp with Silicon-maid-7b.Q5_K_M.gguf using less than 6GB video memory The key requirements are: - video memory as much as you can get - a model with low memory requirements - tools to allocate video memory for game rendering (ENBoost) and LLM inference (KoboldCpp)
When I press key button for "Change backend model" (Advanced Way) nothing happen. It should change to kobaldcpp, but i don't have a message about it and Herika says nothing to my texting.
If anyone is familiar, I need a little help. I sudddenly started getting this error : Warning: fopen(https://api.openai.com/v1/chat/completions): Failed to open stream: HTTP request failed! HTTP/1.1 429 Too Many Requests in /var/www/html/HerikaServer/connector/openai.php on line 115
Fatal error: Uncaught TypeError: fgets(): Argument #1 ($stream) must be of type resource, bool given in /var/www/html/HerikaServer/connector/openai.php:136Stack trace: #0 /var/www/html/HerikaServer/connector/openai.php(136): fgets() #1 /var/www/html/HerikaServer/ui/tests.php(84): connector->process() #2 {main} thrown in /var/www/html/HerikaServer/connector/openai.php on line 136
Its happening since last 2 days. I tried a different account with different API keys, same account with new API keys etc. Same error each time.
You have reached rate limits (tokens per minute, requests per minute or day) or usage limits (monthly spend limits). In openAi platform check "Usage" and current limits (Settings -> Limits) to see where is the problem. Solutions: - reduce prompt size (PROMPT_HEAD, HERIKA_PERS) and memory size (CONTEXT_HISTORY, MAX_TOKENS_MEMORY) and wait at least a day without any new request until rate limit fall in accepted range; - increase usage limits by changing monthly budget. From my experience, with default settings, after filling the memory, Herika prompt is over 1000 tokens per request. With actions enabled and memory filled, she can be very talkative sending enough requests to trigger error 429 for some of the models.
I've been trying to adjust some settings in the Herika server but don't seem to come up with noticable results. I'm using the open ai api for both the ai itself and the tts.
I want to improve her memory, and also give her the ability to tell longer dialogs without getting cut off.
It would also be nice to tweak her so that she doesn't do the lengthy narrative of herself every time = "Herika tell us a story" / "Ok I will tell you a story that will leave your mind bewildering using my incredible story telling skills...*4 paragraphs later about how great the story will be* ... and here it goes, Once..." oh, past the tokens limit some how .. Just tell the damn story!! lol
If anyone has a better idea on some of these settings, or a resource that actually covers all those settings on the server and what they do better it would be much appreciated. I tried adjusing a couple settings that seemed like it might help but I don't want to break it without knowing for sure.
For anyone wanting to route the WSL IP to the official windows LAN IP, check out this thread: https://github.com/microsoft/WSL/issues/4150 There's a powershell script setting up port forwarding and firewall. Works great! Cheers
952 comments
https://discord.gg/NDn9qud2ug
I mean I still get the API with the free version or... i'm wrong?
GPT is super cheap though. I didnt enable the visual aspect, but basically I have run it with quite heavy & lengthy sessions with Mantella running at the same time and its like less than 10 bucks a month for both running off the GPT 3.5.
( :( )
I am already leaving out other issues
I"m running with about 900 mods and Herika. Her avatar is completely compatible with flower girls/romance mods, amazing followers tweaks, dozens of NPC tweaks and modifications, and just about anything else you can possibly think of that would in any way interact with Herika herself.
It is a bit challenging and I haven't come up with a solution to get it working with Mantella the way "I" would like her to. It works, with radiant conversations and everything else, but its like a different persona in Mantella than the Herika we all know. Basically she has multiple personalities between the Mantella Herika and the "regular" Herika. I am hoping to find a way to set her up so that she can partake in radiant confversations with Mantella driven NPCs with the "real" Herika. Other than that, which may not even be possible yet, Herika is compatible with everything. I wanted to add this post incase anyone is considering putting her in the game, so a nonsense post doesn't scare them off.
PS Antother thing I would really like is to be able to set up multiple "herika" servers running simutaniously so that my other followers can have the same robust AI with their own personalities. Again, I don't think this is possible yet but I'm not sure.
*edit - 1 week later* I started with a 13B LLM in an older format, the person in the video I used to get some starting advice recommended using one in a .gguf format even though his links started with the older one. I found a 7B LLM in .gguf, and what a difference. It ran so much better, but not perfect. Pretty sure I would need some upgrades to smooth it out. It was mostly some frame skips if I came near an NPC that I could talk to, and Herika was starting to listen.
- AMD Ryzen 9 5900X
- RAM 64GB
- video RTX 3060 12GB
- ENBoost to extend video memory used by Skyrim
- KoboldCpp with Silicon-maid-7b.Q5_K_M.gguf using less than 6GB video memory
The key requirements are:
- video memory as much as you can get
- a model with low memory requirements
- tools to allocate video memory for game rendering (ENBoost) and LLM inference (KoboldCpp)
I sudddenly started getting this error :
Warning: fopen(https://api.openai.com/v1/chat/completions): Failed to open stream: HTTP request failed! HTTP/1.1 429 Too Many Requests in /var/www/html/HerikaServer/connector/openai.php on line 115
Fatal error: Uncaught TypeError: fgets(): Argument #1 ($stream) must be of type resource, bool given in /var/www/html/HerikaServer/connector/openai.php:136Stack trace:
#0 /var/www/html/HerikaServer/connector/openai.php(136): fgets()
#1 /var/www/html/HerikaServer/ui/tests.php(84): connector->process()
#2 {main}
thrown in /var/www/html/HerikaServer/connector/openai.php on line 136
Its happening since last 2 days. I tried a different account with different API keys, same account with new API keys etc. Same error each time.
Solutions:
- reduce prompt size (PROMPT_HEAD, HERIKA_PERS) and memory size (CONTEXT_HISTORY, MAX_TOKENS_MEMORY) and wait at least a day without any new request until rate limit fall in accepted range;
- increase usage limits by changing monthly budget.
From my experience, with default settings, after filling the memory, Herika prompt is over 1000 tokens per request. With actions enabled and memory filled, she can be very talkative sending enough requests to trigger error 429 for some of the models.
I want to improve her memory, and also give her the ability to tell longer dialogs without getting cut off.
It would also be nice to tweak her so that she doesn't do the lengthy narrative of herself every time = "Herika tell us a story" / "Ok I will tell you a story that will leave your mind bewildering using my incredible story telling skills...*4 paragraphs later about how great the story will be* ... and here it goes, Once..." oh, past the tokens limit some how .. Just tell the damn story!! lol
If anyone has a better idea on some of these settings, or a resource that actually covers all those settings on the server and what they do better it would be much appreciated. I tried adjusing a couple settings that seemed like it might help but I don't want to break it without knowing for sure.
I am a little bit afraid to update to the new version, because of this: "Updated to 1.6.1170".
Does anyone know if this works fine with VR?
https://github.com/microsoft/WSL/issues/4150
There's a powershell script setting up port forwarding and firewall. Works great!
Cheers