Agents

Three related loops on Developer Sandbox. Prefer Golden Paths to create a new agent Pod + catalog Component from real LangGraph / LangChain.js / LangChain4j source.

1. Hub Guest (in-cluster)

Guest has no OpenShift token. Agent Deployments and builds are created by chart RBAC (samples + agent-applier), not by the Guest’s kubeconfig.

2. Agent Pod (Golden Path / samples)

Golden Path (build=true)

• Scaffolder generates framework source; agent-applier starts a Binary BuildConfig.
• Image: image-registry.openshift-image-registry.svc:5000/<ns>/<name>:latest
• Env: AGENT_SPEC, AGENT_TYPE, MODEL, LiteLLM credentials
• Endpoints: /health, /, /v1/chat (plus /v1/graph or /v1/runtime per language)
• Service: ClusterIP only

oc get bc,is,deploy,svc -l app.kubernetes.io/component=agent
oc logs -f bc/<name>
oc logs deploy/<name>

Legacy samples

Pre-installed sample-*-agent Deployments were removed. Create agents only via the Deploy Agent golden path (build=true).

3. DevSpaces (OpenShift user)

Use Golden Path Agent-friendly DevSpaces AI Workspace: pick language → a started DevWorkspace appears with Continue → LiteLLM.

Wire Continue to the LiteLLM Route with litellm-master-key (see DevSpaces AI).

Factory fallback (same language skeletons):

https://workspaces.openshift.com/#https://github.com/maximilianoPizarro/rhdh-agent-sandbox/tree/main/files/templates/skeletons/<language>

Official Red Hat agentic skills (CVE, lifecycle, diagnostics, support severity) are available in the catalog and can be installed into IDE skill directories.

Identity cheat sheet