Using Downsizing with Dify

Prerequisites

A running Dify workspace (Cloud or self-hosted) with permission to manage Model Providers, 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

Dify's built-in Anthropic provider doesn't expose a base URL/endpoint override, so it can't be pointed at a proxy. Dify does ship an OpenAI-API-compatible provider for exactly this purpose — use it to route through Downsizing:

1. In your workspace, go to Settings → Model Provider (or Integrations → Model Provider depending on your Dify version).
2. Find OpenAI-API-compatible and click Add Model.
3. Fill in the form:
   • Model Type: LLM.
   • Model Name: any label, e.g. a Claude or GPT model name.
   • API Key: your Downsizing API key (starts with dws_).
   • API endpoint URL: your Downsizing OpenAI-compatible base path (ending in /v1, without the /chat/completions suffix):

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

Replace YOUR_INFERENCE_ID with your workspace UUID. Save, then set this model as the default LLM for your app under Model Settings.

Verify

Open a chatflow or the Playground and send a test message. If you get a response, the provider is wired up correctly. Then check your savings dashboard to see the request appear in your usage metrics.