Using Downsizing with SillyTavern

Prerequisites

A running SillyTavern instance and a Downsizing account that can make requests (pay-per-use or a connected subscription both work). Grab your Inference ID and API key from the Inference API page.

Manual Setup

SillyTavern's built-in Claude Chat Completion source only takes an API key — it has no field for a custom endpoint. The Custom (OpenAI-compatible) source does support a custom endpoint, so use that to route through Downsizing:

1. Open the API Connections panel and set API to Chat Completion.
2. Set Chat Completion Source to Custom (OpenAI-compatible).
3. Fill in:
   • Custom Endpoint (Base URL): your Downsizing OpenAI-compatible endpoint, ending in /v1 (do not add a /chat/completions suffix):

http://localhost:3001/YOUR_INFERENCE_ID/openai/v1

• Custom API Key: your Downsizing API key (starts with dws_).
• Custom Model ID: pick a model from the dropdown if it loads from /v1/models, otherwise type one manually (e.g. a Claude model name).

Replace YOUR_INFERENCE_ID with your workspace UUID from the dashboard. Click Test Message to confirm the connection; if SillyTavern reports a false-negative status warning, enable Bypass status check.

Verify

Send a message in any chat and confirm you get a reply. Then check your savings dashboard to see the request appear in your usage metrics.