itch.io

Since you've edited your post while I was writing mine:

The description of your custom character (or the name of one of your attached facial expressions) probably contains some kind of letter that breaks JSON formatting or OpenRouter just doesn't like somehow.

Maybe you have "ZDR Endpoints Only" enabled in your OpenRouter settings? If you've deleted the settings file and are still getting this error with an untouched Quick Setup, I can only imagine this being a problem on OpenRouter's end. Sometimes providers also just go offline, but I can confirm that the one Quick Setup uses is currently working.

Are you trying to launch the game via koboldcpp.exe? You have to start Silverpine.exe instead.

There doesn't really seem to be a problem here. What exactly happens when you try talking to an NPC?

Please post your GPU, a screenshot of the error if possible, and the name of the model you're trying to load.

There are pretty clear instructions/a visual guide/live previews/examples in the custom player character UI. You simply apply the expression to the base portrait in your image editing software of choice, crop the image vertically so it only includes the changed part, then import it into the game.

The expressions apply seamlessly to your nude alt, as long as there is no vertical overlap between the face and the clothing like this:

I think you floated the idea of allowing us to substitute alternate art for them at some point, so this is kind of brand. 

This will come in the form of custom NPCs. Custom NPCs will have several hardcoded template behaviors you can choose from (and perhaps custom behaviors in the form of Lua scripts or some kind of simple visual scripting solution), one of which will be taking the place of an existing NPC.

Report bugs here. Ideally, your report should include the steps necessary to reproduce the bug.

Post game related feedback here.

I didn't realize you were talking about the smaller Gemma that no longer ships with the game, which is still available to you because it exists as a custom model definition.

In that case, you should use Nemo instead. It will run at about the same speed as the smaller Gemma.

For advanced setups like this that are not supported by the game, you must start KoboldCPP manually with your desired parameters, and then start the game, which will automatically detect the running process and start communicating with it. This is only available on Windows. The game will assume the process it's hooking into is running the same model last selected after the "Select an AI model for more information." dialog.

How much VRAM do you have? If it's 12 GB or 16 GB, I highly recommend using Mistral-Small-3.2 instead. Before, there were manual offload values for Gemma that were very much guestimates, which I got rid off, so the game now asks the backend to determine the correct amount, which it probably fails to do for some reason in your case.

Change this value in AppData\LocalLow\Three Eyes Software\Silverpine\settings.json to anything else, and the game will interpret it as your GPU having changed, showing you the dialog again.

Read this post. There should be an e-mail somewhere in your inbox.

Please read the post again carefully:

If you bought the game before the time of this post

The whole reason I made this post is because hosting the server outside of the demo turned out to not be sustainable.
I'm merely honoring the expectation I created for people who bought the game before it. Please run the AI locally instead. If you can't run it locally, use OpenRouter.

Judging from the resolution of your screenshot, your system doesn't have the VRAM required to run any of the local models. Use OpenRouter instead.

The game shows you an error that tells you what exactly failed during the conversion. If your character is called "Example Custom Character" it will also not be converted.

Alternatively, you can simply move the folder of your custom player character definition somewhere else, and then manually recreate it in the new version's editor using the assets inside.

If the curl command fails, the issue of connecting to localhost is related to something else on your system, not the game itself. Perhaps there's another program already running on port 5001.

As for the slowness, the game is correctly offloading all layers to the GPU. I can't fully gauge the performance from the two successful API calls you posted since they have very few input/output tokens, but I've uploaded a new version that makes 5000 series RTX GPUs use a special backend setting, which when not selected, resulted in much slower (though not minutes long) processing during my testing.

I've uploaded updated builds where the game will let you copy the error message to your clipboard. I imagine it's related to a buggy item in your inventory.

What version did this first happen on? Does it happen on fresh saves?

Open a command prompt in Silverpine_Data\StreamingAssets\KoboldCPP like this:

Then run this command:

koboldcpp.exe --model "Mistral-Small-3.2.gguf" --usecublas --gpulayers 999 --quiet --multiuser 100 --contextsize 4096 --skiplauncher

Then post this part of the output:

Then run this command in a separate command prompt:

curl -X POST "http://localhost:5001/api/v1/generate" -H "accept: application/json" -H "Content-Type: application/json" -d "{\"max_context_length\": 4096,\"max_length\": 100,\"prompt\": \"Lorem ipsum dolor sit amet, consectetur adipiscing elit. Mauris laoreet nunc non vehicula accumsan. Etiam lacus nulla, malesuada nec ullamcorper vitae, malesuada eget elit. Cras vehicula tortor mauris, vitae vulputate est fringilla ac. Aenean urna libero, egestas eget tristique eget, tincidunt sit amet turpis. Pellentesque vitae nulla vitae metus mattis pulvinar. Suspendisse eu gravida magna. Nam metus diam, fermentum mattis pretium vestibulum, mollis non sem. Etiam hendrerit pharetra risus, vitae fermentum felis hendrerit at. \",\"quiet\": false,\"rep_pen\": 1.1,\"rep_pen_range\": 256,\"rep_pen_slope\": 1,\"temperature\": 0.5,\"tfs\": 1,\"top_a\": 0,\"top_k\": 100,\"top_p\": 0.9,\"typical\": 1}"

After a while something like this should pop up in the first command prompt:

Please post it too. I should be able to figure out the issue then.

It's probably your firewall or something else on your network if both the server and OpenRouter aren't working.

It shouldn't take more than a few seconds. Sounds like the game can't detect your VRAM amount correctly. Are you on Windows or Linux? Do you have an iGPU alongside your 5070ti (e.g. laptop)?

Should be fixed now. Just redownload.

NPCs should usually end the dialog and start following you in any kind of "going somewhere" situation, but I will investigate adding this as a feature.

You're right, it's not getting mixed. Thanks for the report.

It's been more difficult because I did indeed change something about the way this is checked. I might revert the change, since people seem to prefer actions sometimes happening inappropriately over actions sometimes not happening when appropriate.

NovelAI doesn't seem to support proper prefill outside of explicitly setting and then unsetting the prefill as 2 additional separate API calls, which would lead to race conditions, so not for now.

https://itch.io/post/14837950/

After trying this on Arch, it turned out to be a simple problem caused by Arch being more strict with file permissions.

It can be solved via chmod +x koboldcpp-linux, but I've also uploaded a new version where this happens automatically.

Well it definitely works on Linux Mint 22.2, but the game uses Linux's popen() function instead of the system() function (which i believe the terminal uses) to spawn the process, so I'll add that as a fallback tomorrow. What distro are you using?

Diagnostic.bat is included by mistake.

Really you're not meant to edit the exported NPC memories at all, and I can't imagine editing them not causing confusion in some way. NPCs receive information by being part of a conversation, or by having information shared with them by another NPC.

I see that people like having overarching plotlines, so I'd rather have a system that handles this better, than having people edit the exported memories. I'm very much open for suggestions about this if you feel like there's something missing, or that maybe there should be more detailed/frequent/reliable communication.

The field that contains the used actions does nothing if you don't have "Native NPC Actions" enabled. Otherwise it tells the AI what actions were already used. Again, it's not meant to be edited.

For Nemo, it should take no more than 2-3 seconds on a 5070 ti, but I actually worked on porting the game over to Linux today (which also uses a linux native backend for the LLM), so try the native build.

I've sent you an e-mail.

What service is this from? It looks like something split up the model into parts. The game currently only supports models with a single submesh/material.

I have found a solution for this now, but I probably won't be able to upload the patch until tomorrow.