Apps
SillyTavern setup
SillyTavern talks to any OpenAI-compatible API. Point its Custom source at nokeep, test with the free key, then chat with an uncensored model that stays in character.
Checked 26 October 2026
The values
- Custom Endpoint (Base URL)
https://nokeep.ai/v1- API key
sk-nokeep-test to test, then your own sk-nokeep-… key- Model
deepseek-v4.1-flash-uncensored
Start with the free test key: it answers one fixed line, which proves the setup is right without spending anything. Then paste your own key from the console. deepseek-v4.1-flash-uncensored for most chats and pictures, or glm-5.3-flash-uncensored for text at under half the price.
Connect
- Open API Connections: the plug icon in the top bar.
- Set API to Chat Completion.
- Set Chat Completion Source to Custom (OpenAI-compatible).
- Paste
https://nokeep.ai/v1into Custom Endpoint (Base URL). Stop at/v1: SillyTavern adds the rest. - Paste
sk-nokeep-testinto Custom API Key. - Press Connect. The model list fills from our
/v1/models; pickdeepseek-v4.1-flash-uncensored, or type it as the model id. - Press Test Message. A reply that starts “Connected to” means everything is right.
- Swap in your own key from the console and press Connect again. Save it as a connection profile if you like.
Labels can move between SillyTavern versions. If one isn’t where this says, look in the same panel.
Reply length and samplers
DeepSeek answers at once unless asked to think, so SillyTavern’s usual reply length of about 300 tokens gives a whole paragraph. Raise it for longer replies.
DeepSeek takes Temperature, Top P, Top K, Min P and the Frequency, Presence and Repetition penalties from SillyTavern’s sliders. Good starting values are on Samplers for fiction. GLM ignores the sliders and uses its own settings.
Thinking, if you want it
To have DeepSeek think before it answers, add a body field in the connection panel’s Additional Parameters, then raise the reply length to 1,000 tokens or more so the thinking doesn’t use it all.
reasoning_effort: highIf it doesn’t connect
| The app shows | Fix |
|---|---|
| 401 | The key is wrong or revoked. Paste it again, with no spaces or quotes. |
| 404 | The URL has the path twice. Use https://nokeep.ai/v1, without /chat/completions. |
| 402 | Your account is out of bought credits. Top up in the console, from $5. |
| 429 | Over 120 requests a minute, or the key’s daily cap. Raise the cap in the console, or wait. |
| 503 | The model is busy. Send again in a few seconds; nothing was charged. |
| The fixed test line | The setup works. Swap in your own key to get real replies. |
Every error is listed with its fix in Errors.
What an evening costs
At 3,000 tokens in and 250 out a reply, $5 is about 2,700+ replies on DeepSeek, or 6,500+ replies on GLM. A longer character card or a longer chat history costs more per reply, since the whole history is sent each time. Work out your own.
Questions
Does nokeep add a prompt to my character card?
No. SillyTavern’s prompt is everything the model sees. Nothing of ours is added, and nothing is filtered out of the reply.
Do I need a jailbreak prompt?
No. Both models have their refusals removed from the weights, so they answer without one.
Where is my chat stored?
In your SillyTavern, on your machine. We keep one row of counts per call, never the messages.