#
SillyTavern
Step-by-step instructions on how to use Mancer with SillyTavern.
#
Step 0 - Make a Mancer account
Click here. Go to the site, do the thing, profit. You can use some models completely for free, but if you pay us you can get access to fancier models.
#
Step 1 - Install SillyTavern
You can get it from their github.
They have incredible documentation, use it. (Note the 'Installation' category on the side!)
#
Step 2 - Launch SillyTavern
If you can't get ST to run, this is not the place to find help.
#
Step 3 - Set the API connection
#
HELP! - I can't see a 'Mancer AI' checkbox!
Update your installation of SillyTavern, it's probably pretty old. Maybe back up your chats/cards first, though.
#
Step 4 - Set your API key
You can find your API key on your dashboard. Don't share this key around, it's all anyone needs to spend your credits. You can reset your API key from the Dashboard.
#
Step 5 - Pick a model
Take a look at this live-updated list of all models!
You can switch models any time you want just by changing the URLs you set here, but remember that different models have different credit costs and prompting/settings expectations.
#
Step 6 - Try to connect
#
HELP! - I can't connect! (SillyTavern yelled at me with a popup!)
That popup should tell you what went wrong, including:
- Wrong API key: Make sure you copied yours in correctly!
- Bad model name/url: Make sure you copied it in correctly!
#
HELP! - I can't connect! (No connection...
)
- Make sure you copied in the URL correctly (it should start with https, not http!)
- Make sure you have internet. This is not a "local" service, you need to be able to connect to us.
- Check if you can access our website. If that's down, we have bigger problems.
- If you can't access the website (how are you reading this?), then you may have a DNS issue, which is waaaay out of scope to address here.
- If no one else can access the website either, then, uh...
#
Step 6 - Generate a response
Pick a card, send a message, see if you get something back.
#
HELP! - I'm getting terrible replies!
Welcome to the land of custom language models.
Take a look at the documentation pages for whatever model you're trying to use. Each one might require different settings to get the most out of it. Commonly, that means different instruct-prompt values, different sampling parameters, or better card greetings/example messages.
There's an entire world of things that can affect what a language model will spit back at you. Everything from the previous content of your chatlog to SillyTavern's own settings. It might take some work, but you'll end up with far more control over what kinds of responses you get than ever before.