Setup guides
The Roleplay API speaks OpenAI chat completions, so any app with an OpenAI-compatible mode can use it. One key works in all of them. Pick your app and paste three things: the URL, your key and a model id from Models.
- Get a key. Sign in, open the API dashboard, name a key and copy it.
- Paste the URL. Each app wants it in a slightly different form, listed below.
- Pick a model. Type an id such as
hanami-70b-x1, or let the app fetch the list.
JanitorAI
API Settings, Proxy. Paste the URL and your key, and type a model id.
https://kyrune.com/api/v1/chat/completionsThe shorter https://kyrune.com/api/v1 form is accepted too.
SillyTavern
API: Chat Completion. Source: Custom (OpenAI-compatible). Paste the URL and key, then Connect.
https://kyrune.com/api/v1Do not add /chat/completions to the URL.
Chub
Secrets, OpenAI, Reverse Proxy. Enter the URL, your key and a model id.
https://kyrune.com/api/v1Agnai
Self-Host, OpenAI Compatible, OpenAI Compat (Chat).
https://kyrune.com/apiLeave "No URL suffix" off so Agnai adds /v1 itself.
RisuAI
OpenAI Compatible with URL autofill on.
https://kyrune.com/api/v1With autofill off, use https://kyrune.com/api/v1/chat/completions.
Open WebUI
Connections, OpenAI-compatible API base URL.
https://kyrune.com/api/v1LibreChat
A custom endpoint with this baseURL.
https://kyrune.com/api/v1Model fetching can use /models.
LobeChat
A custom OpenAI-compatible base URL.
https://kyrune.com/api/v1AnythingLLM
Generic OpenAI, base path.
https://kyrune.com/api/v1Chatbox
Custom OpenAI API host.
https://kyrune.com/api/v1Chatbox also accepts https://kyrune.com/api and adds the version itself.
NextChat
OpenAI endpoint root.
https://kyrune.com/apiDo not add /v1: NextChat appends v1/chat/completions itself.
OpenAI SDKs
Set it as baseURL and send your key as a Bearer token.
https://kyrune.com/api/v1Normal and -raw models
A plain model id uses the preset tuned in your Kyrune dashboard, so an app's hidden sampler defaults cannot change how it writes. Add -raw to the id and the sampler values your app sends pass straight through. stream and stop always come from the request.
If something fails
| You see | What it means |
|---|---|
401 | The key is missing, wrong or revoked. |
404 | The URL path is wrong or doubled, for example /v1/v1. |
400 unknown model | The saved model id is not in the current model list. |
429 | Your plan's daily limit or stream limit is reached. |
An error during a reply | The request was accepted and failed partway through the stream. |
If it worked before you pressed Save in your app and fails after, open the connection again and check the URL, key and model survived, without a doubled suffix.