Skip to content

SillyTavern ​

A chat and roleplay front end. Use the Custom (OpenAI-compatible) source.

Connect ​

API Connections → API: Chat Completion → Chat Completion Source: Custom (OpenAI-compatible):

  • Endpoint: http://127.0.0.1:4010/v1 (don't add /chat/completions)
  • API key: your plugboard key
  • Model: picked from the dropdown, which fills from /v1/models

Press Test Message to confirm.

In Docker, use http://host.docker.internal:4010/v1 as the endpoint. The compose file maps that name to the host (examples/sillytavern/):

yml
services:
  sillytavern:
    image: ghcr.io/sillytavern/sillytavern:latest
    ports:
      - "8000:8000"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    volumes:
      - ./config:/home/node/app/config
      - ./data:/home/node/app/data

What works ​

Chat with streaming, and tool calls when prompt post-processing is None or a "with tools" variant.

Limits ​

  • The Strict and Single user message post-processing options merge roles; you don't need them with plugboard.