Hermes Atlas
Deployment & hosting

Hermes Stack

Godde3s/hermes-stack

Hermes Agent, 9Router and OmniRouter on a free Hugging Face Space with a deploy wizard and hourly backups

In short

Hermes Stack is a deploy template that runs Hermes Agent, 9Router and OmniRouter on a single free Hugging Face Space. It adds a web dashboard, an OpenAI-compatible API, a Telegram gateway, hourly backups and a keep-alive action.

What Hermes Stack does

Hermes Stack packages Hermes Agent, 9Router and OmniRouter into one deployment on a free Hugging Face Space, which the README describes as 2 vCPU and 16 GB RAM with no credit card. A web wizard, a GitHub template or a manual huggingface-cli route creates the Space, and the first build takes roughly 15 to 25 minutes. A single Caddy front proxy serves a router dashboard, an OpenAI-compatible API, the agent's web dashboard behind basic auth, and the agent itself as an OpenAI-compatible API.

Because Spaces have ephemeral disks, the stack tarballs the whole agent state hourly to a private Hugging Face dataset and restores it on restart. A GitHub Action pings a health endpoint every 10 minutes to keep the Space awake, a supervisor restarts crashed services with exponential backoff, and a Telegram gateway lets you chat with the agent. OmniRouter serves about 78 free web models without API keys. Extra packages and a persistent boot script can be added through environment variables.

Key features

  • Web deploy wizard that validates GitHub, Hugging Face and Telegram tokens and starts the deploy
  • Single Caddy proxy exposing a router dashboard, an OpenAI-compatible API and the agent dashboard
  • Telegram gateway that uses polling, so it works behind a firewall
  • Hourly backup of the agent state to a private Hugging Face dataset with automatic restore
  • Keep-alive GitHub Action that pings /healthz every 10 minutes
  • Self-healing supervisor with exponential backoff

When to use it

  • Stand up a personal Hermes Agent server without paying for a VPS
  • Expose the agent as an OpenAI-compatible API for other clients
  • Rebuild the whole agent state into a fresh Space from the backup dataset

Who it is for: Individuals who want a hosted Hermes Agent with a web dashboard and Telegram access on a free Hugging Face Space.

How it fits with Hermes Agent

Built around Hermes Agent as a deploy template, with 9Router and OmniRouter supplying the model routing and the agent exposed as its own API.

Requirements: A GitHub account and token, a Hugging Face write token and username, and optionally a Telegram bot token and your numeric Telegram ID

Note: Free Hugging Face Spaces use ephemeral disks and sleep after inactivity, so the stack depends on hourly backups and a keep-alive GitHub Action.

FAQ

What is Hermes Stack?

Hermes Stack is a one-click deploy template that runs Hermes Agent, 9Router and OmniRouter together on a free Hugging Face Space. It adds a web dashboard, an OpenAI-compatible API, Telegram control, hourly backups and a keep-alive job.

Does Hermes Stack work with Hermes Agent?

Yes, Hermes Agent is the core component. It runs behind the stack's front proxy, and its web dashboard and an OpenAI-compatible API for the agent are exposed under /hermes/ and /hermes-api/v1.

How do I deploy Hermes Stack?

There are three ways: the web deploy wizard, a GitHub template plus Actions secrets followed by the Deploy to Hugging Face Space workflow, or a manual huggingface-cli deploy. The first build takes about 15 to 25 minutes.

Similar deployment for Hermes Agent

All deployment

Related guides: How to run Hermes Agent securely · How to install Hermes Agent · Connect Hermes agents on several machines with Hermes Desktop