NovaAPI

English

简体中文

English

简体中文

Appearance

Hermes ​

Hermes (Hermes Agent by Nous Research) is OpenAI-compatible. Add NovaAPI as a custom endpoint and swap the key for a NovaAPI Key. This guide starts from scratch and covers Windows / macOS / Linux.

Requirements ​

  • OS: Windows 10+ (PowerShell) / macOS / Linux (WSL2 also works)
  • An API Key from the NovaAPI console (starts with sk-)

No manual Node.js/Python setup needed — the Hermes installer bundles its own runtime (uv, Python 3.11, Node.js, ripgrep, ffmpeg). Install only what you're missing.

1. Install Hermes ​

The official installer detects your platform and pulls the runtime it needs.

macOS / Linux / WSL2 ​

bash
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

After it finishes, reload your shell so hermes is on PATH:

bash
source ~/.bashrc      # or: source ~/.zshrc

Windows (PowerShell) ​

powershell
iex (irm https://hermes-agent.nousresearch.com/install.ps1)

Reopen the terminal afterward.

Verify:

bash
hermes --version

2. Point it at NovaAPI ​

NovaAPI is an OpenAI-compatible endpoint. Two ways to configure it.

bash
hermes model

Choose "Custom endpoint (self-hosted / VLLM / etc.)", then enter:

  • API base URL: https://api.novaapis.com/v1
  • API key: your sk-... NovaAPI key
  • Model name: any tool-calling model from the model list

Your choice is saved to ~/.hermes/config.yaml.

Option B — edit the config manually ​

Add NovaAPI as a named custom provider in ~/.hermes/config.yaml, and put the key in ~/.hermes/.env:

yaml
# ~/.hermes/config.yaml
custom_providers:
  - name: novaapi
    base_url: https://api.novaapis.com/v1
    key_env: NOVA_API_KEY

model:
  default: gpt-5            # any tool-calling model
  provider:
    custom: novaapi
bash
# ~/.hermes/.env
NOVA_API_KEY=sk-your-NovaAPI-key
  • base_url: NovaAPI's OpenAI-compatible endpoint
  • key_env: the env var in .env that holds the key
  • model.provider.custom: the custom provider name defined above

config.yaml is the single source of truth for model, provider, and base URL.

3. Run ​

bash
hermes

Switch models mid-session with /model, or re-run hermes model to reconfigure. Run hermes doctor if a provider isn't recognized.