Using Downsizing with Hermes

Prerequisites

Install the Downsizing CLI — see the CLI guide if you haven't already.

CLI Setup (Recommended)

The easiest way to connect Hermes is with the Downsizing CLI. If you haven't installed it yet, grab it from the CLI guide.

Sign in, then link Hermes:

ds link hermes

The CLI writes the correct base_url and api_key to ~/.hermes/config.yaml. The command is idempotent — safe to run multiple times.

Manual Setup

If you prefer to configure manually, edit ~/.hermes/config.yaml and set:

model:
  default: claude-opus-4-8
  provider: custom
  api_mode: codex_responses
  base_url: http://localhost:3001/YOUR_INFERENCE_ID/openai/v1
  api_key: dws_YOUR_API_KEY

Find your Inference ID and API key in the Inference API page. Replace YOUR_INFERENCE_ID with the workspace UUID and dws_YOUR_API_KEY with your Downsizing API key.

How it works

Hermes sends all model requests through Downsizing over the OpenAI Responses API, using your Downsizing API key for authentication. The proxy validates the request, applies routing and reduction, then forwards the call using your stored credentials. Responses stream back to Hermes unchanged.

Verify

Run a quick one-shot query to confirm Downsizing is working:

hermes chat -q "test"

If Hermes responds, Downsizing is set up correctly. You can also check your savings dashboard to see the request appear in your usage metrics.

Unlinking

To remove Downsizing from Hermes:

ds unlink hermes