Skip to content

Getting Started

Use this guide to bring up Hermes Agent locally and confirm that the gateway and dashboard are reachable.

Prerequisites

  • Docker with Compose support
  • kubectl
  • curl
  • kind, available at tools/kind
  • A Z.AI GLM Coding Plan API key when you want live model calls

The repository keeps generated runtime state out of Git. Compose data is written to data/, and the kind helper binary is expected at tools/kind.

Clone and Prepare

bash
git clone https://github.com/Sunwood-ai-labs/hermes-agent-pod.git
cd hermes-agent-pod

For interactive Hermes setup through Docker:

bash
./scripts/setup.sh

Start the kind Pod

bash
./scripts/kind-up.sh

The script creates or reuses the sandbox-hermes kind cluster, applies the sandbox-hermes namespace, preserves an existing hermes-secrets Secret when present, and waits for the hermes-agent Pod to become ready.

To configure a GLM Coding Plan key before or after starting the Pod:

bash
GLM_API_KEY="..." ./scripts/set-glm-key.sh

Then verify the Pod:

bash
./scripts/kind-verify.sh

Start with Docker Compose

Use Compose when you want the quickest fallback without a Kubernetes Pod:

bash
./scripts/up.sh
./scripts/verify.sh

Stop the Compose service with:

bash
./scripts/down.sh

Stop the kind Runtime

bash
./scripts/kind-down.sh

This deletes the sandbox-hermes kind cluster. Because the Pod data lives in a cluster PVC, deleting the cluster also removes that Pod-side state.

Released under the MIT License.