Skip to content

Use ZhiFlo in Hermes

Hermes provides a Custom endpoint flow for the ZhiFlo address, Token, protocol, and model. After you save the provider, hermes chat can use it for a real reply.

  • Install Hermes from its official documentation, then confirm that hermes --version prints version information.
  • Create a Token for this device.
  • Copy a complete model ID from your account.

On Windows, search for PowerShell from Start. On macOS, open Terminal; on Linux, open a terminal from the application menu. If the command is missing immediately after installation, close the old terminal, open a new one, and check the version again.

  1. Exit the current Hermes chat if one is open.

  2. Run this command in a normal terminal:

    Terminal window
    hermes model
  3. Select Custom endpoint.

  4. Enter this API Base URL:

    https://api.zhiflo.com/v1
  5. Paste the dedicated ZhiFlo Token at the API Key prompt.

    Paste the Token only at this credential prompt. Keep it out of commands, screenshots, and chat messages.

  6. Select this compatibility mode:

    chat_completions
  7. Select an account model from the discovered list. If discovery does not return it, enter the complete model ID manually.

  8. Review the address, compatibility mode, and model ID in the summary.

  9. Save the provider.

Use hermes model to add a provider. The /model command inside a chat only switches among providers that have already been configured.

Run:

Terminal window
hermes chat

Confirm that the current provider or model points to the ZhiFlo entry, then send:

Do not call tools. Reply with OK only.

An OK reply shows that the provider was saved, the model was selected, and a request returned content.

  • Model discovery is empty: enter the complete model ID from your account manually. Discovery failure alone does not prove the entered model is unavailable.
  • ZhiFlo is absent from /model: exit the chat, rerun hermes model, and complete the final save step.
  • 401 response: reopen the setup flow and update the Token, checking for missing characters, expiration, or revocation.
  • 404 response: use https://api.zhiflo.com/v1 with chat_completions.

Next: Review usage and quota.