OpenClaw (AI Agentic Platform)
OpenClaw is an AI-powered agentic platform designed to facilitate autonomous task execution and browser automation.
1. Overview
OpenClaw is deployed as a containerized application within the K3s cluster, serving as the primary interface for AI agent interactions.
- Status: Active
- Target Environment: Cloud (Oracle Cloud)
2. Requirements & Prerequisites
Necessary components or states prior to implementation:
- Helm Controller: Installed in the K3s cluster.
- Sealed Secrets:
openclaw-credentialscontaining OpenRouter API keys. - Access Control: Tailscale or Cloudflare Tunnel for connectivity.
3. Implementation Procedure
The service is managed through ArgoCD using the official OpenClaw Helm chart.
A. Environment Configuration
- Namespace Creation: The
openclawnamespace is initialized. - Secret Management: Sensitive credentials are encrypted using Bitnami Sealed Secrets.
B. Service Deployment
OpenClaw is deployed to the k3s-master node to utilize optimized CPU resources.
4. Configuration Standards
Settings are managed according to the following standards:
- Persistence: A 5Gi volume is provisioned using the
local-pathstorage class. - Model Configuration: The platform is configured to interface with OpenRouter utilizing the kimi-k2.5 model.
- Failover Strategy: Connectivity is maintained with an external Ollama instance on a Windows PC to offload high-compute tasks from OCI.
- Browser Automation: A Chromium sidecar (
alpine-chrome) is enabled for web-based agent tasks. - Resource Limits: The container is restricted to 4GB of RAM to maintain cluster stability.
5. Verification
System health is validated through these procedures:
- Pod Readiness: Verify all containers in the
openclawnamespace are inRunningstate. - Log Inspection: Monitor application logs for successful connection to OpenRouter.
- Accessibility: Confirm the UI is reachable via the configured NodePort or Tunnel.