Hermes

Hermes can use a custom endpoint as the main model provider. Configure the endpoint as OpenAI-compatible Chat Completions, then select the Voxvey model.

1. Set the token

export VOXVEY_TOKEN="<access-token>"

2. Add Voxvey in the dashboard

1. Open the Hermes dashboard.
2. Go to Keys.
3. Add a custom endpoint provider.
4. Use Voxvey as the display name.
5. Use https://api.voxvey.com/v1 as the base URL.
6. Use your Voxvey bearer token as the API key.
7. Set the API mode to chat_completions.

3. Select the main model

Open Models, click Change on the main model row, and select:

deepseek/deepseek-v4-flash

Hermes applies model changes to new sessions. Use /model inside a running chat if you need to switch the current session immediately.

Direct config

You can also edit ~/.hermes/config.yaml.

model:
  provider: voxvey
  default: deepseek/deepseek-v4-flash
  base_url: https://api.voxvey.com/v1
  api_mode: chat_completions
  api_key: ${VOXVEY_TOKEN}

If Hermes stores secrets separately on your install, add the API key through the dashboard or hermes setup, then keep api_key empty in the config file.

Verify

hermes status
hermes chat

In the chat, send:

Reply with the exact word ready.

If Hermes reports no authenticated providers, re-open Keys and confirm the custom endpoint has both a base URL and a token.

Messages mode

The setup above uses Chat Completions. If your Hermes build exposes an Anthropic Messages API mode, use the host root and a model returned by /v1/models for that API surface:

model:
  provider: voxvey
  default: <messages-model-id>
  base_url: https://api.voxvey.com
  api_mode: messages
  api_key: ${VOXVEY_TOKEN}

Use this mode when Hermes should call /v1/messages instead of /v1/chat/completions.