SillyTavern with a custom OpenAI-compatible endpoint

SillyTavern's Chat Completion source accepts a custom endpoint. Point it at the gateway and pick an uncensored or abliterated model id.

Base URL

https://uncensoredai.dev/v1

01
Select Chat Completion
Open the API connection panel, choose Chat Completion as the source and Custom as the provider.

Fill the endpoint

Enter the base URL, key and an uncensored model id.

Connection

Endpoint:  https://uncensoredai.dev/v1
API Key:   sk-ua-XXXX-XXXX-XXXX-XXXX
Model:     glm-5.3-flash-uncensored

03
Tune the context
Set the context size to match the model's window so history is not silently cut. Uncensored GLM tiers support 1M tokens.

Commands and configuration below are shown exactly as the tool expects them, so they stay in English.

Which models suit roleplay?+

GLM-5.3 Uncensored and GLM-5.3 Flash Uncensored are the intended tiers; DeepSeek V4 Pro Uncensored is the alternative.

Are character cards sent to the server?+

Only as part of the prompt you send, exactly like any chat completion request.