Skip to content

Hermes Agent

Hermes Agent

Category: AI Type: Workload Template Tags: llm · ai-agent · hermes · openai-compatible · multi-provider · webui · workload

Overview

Hermes Agent is an AI agent workload with multi-provider LLM support and a built-in WebUI. It supports OpenRouter, OpenAI, Anthropic, and Ollama as backend providers, making it a flexible self-hosted AI assistant that your team can deploy directly within the Juno platform. Each Hermes workload gets its own persistent storage for conversation history and configuration, and can optionally be granted Kubernetes RBAC access to interact with cluster resources.

How It Works

Workload Template — Installs the Hermes Agent workload schema into Genesis. Once installed, the Hermes Agent type appears in Genesis on the Workloads page, where it can be authored into a workload template. Users can then launch and provision their own agent instance on demand within a project through Hubble.

Prerequisites

  • API credentials for at least one supported LLM provider (OpenRouter, OpenAI, Anthropic, or a running Ollama instance)
  • A Kubernetes storage class available in the cluster (for agent state persistence)
  • No install-time prerequisites beyond a running Juno platform

Installation

  1. Open Terra and navigate to the Plugin Marketplace
  2. Search for "hermes-agent"
  3. Click Install
  4. Click Confirm to deploy (no install-time fields required)

Once installed, the Hermes Agent schema is available in Genesis. From the Workloads page, author the template — users can then launch and provision agent instances on demand through Hubble.

Configuration

Install-Time Fields

No install-time configuration is required for this plugin.

Workload Launch Fields

These fields are configured when authoring the workload template in Genesis and used each time a user provisions an agent instance through Hubble:

Field Details
registry string · Optional · Default: nousresearch
Container registry for the Hermes image
repo string · Optional · Default: hermes-agent
Hermes image repository
tag string · Optional · Default: latest
Hermes image tag
cluster_access select · Optional · Default: none
Kubernetes RBAC access level for the agent (none, readonly-ns, admin-ns)
persistence_size string · Optional · Default: 10Gi
Persistent volume size for agent state storage
persistence_storage_class k8sStorageClass · Optional
Storage class for the persistent volume
termination_grace_period int · Optional · Default: 120
Seconds Kubernetes waits for a graceful shutdown (gateway drain + DB checkpoint) before force-killing. Raise on spot/preemptible nodes

Custom Environment Variables

Genesis lets you add arbitrary environment variables to the workload at launch time. These are commonly useful for Hermes Agent's multi-provider setup:

Variable Description
ANTHROPIC_API_KEY Anthropic Console API key, used when the agent is configured to use Claude models.
OPENAI_API_KEY API key for OpenAI or any OpenAI-compatible endpoint (also used for local Ollama setups).
OPENROUTER_API_KEY OpenRouter API key for flexible access to many hosted models.
HERMES_MODEL Overrides the default model used by the agent.

Resilience

Hermes Agent is built to survive unclean node loss (AWS spot-instance reclaims, preemptible nodes, OOM kills):

  • Graceful shutdown — on pod termination, a preStop hook (backed by a signal trap in the init script) drains the gateway, finishes any in-flight turn, and forces a final WAL checkpoint so the SQLite state database is never left mid-write. Tune the window with the termination_grace_period launch field.
  • Startup self-heal — before the gateway opens the state database, an integrity check runs. If the database is corrupt (e.g. after a hard node loss), the recoverable rows are salvaged into a fresh database automatically. Only if the salvage fails is the corrupt original kept (as state.db.corrupt.<timestamp>) for manual repair — the 2 newest such backups are retained, older ones pruned so repeated corruption cannot fill the data volume.

Notes

  • The cluster_access field controls whether Hermes can interact with Kubernetes resources — use readonly-ns for safe exploration, admin-ns only when the agent needs to manage workloads
  • LLM provider API keys are not set via the install-time fields above — configure them either as custom environment variables at workload launch (see table above) or at the application level within the Hermes WebUI after launch
  • Hermes supports OpenAI-compatible APIs, making it compatible with any provider that follows the OpenAI API specification
plugins/hermes-agent/terra.yaml
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
resource_id: hermes-agent
name: hermes-agent
icon: https://github.com/juno-fx/Terra-Official-Plugins/blob/main/plugins/hermes-agent/scripts/assets/hermes-logo.png?raw=true
description: Hermes AI agent with multi-provider support (OpenRouter, OpenAI, Anthropic, Ollama) and WebUI.
tags:
- llm
- ai-agent
- hermes
- openai-compatible
- multi-provider
- webui
- workload
- cluster-level
fields: []