Golden Path journey
End-to-end walkthrough of the three Golden Paths: Deploy Agent, DevSpaces AI Workspace, and AI Service with MCP wiring. Install the chart first via the Helm install journey.
agentType, agentSpec, and LiteLLM model. Framework is mapped deterministically.
catalog:write with build=true → applier hint. No git push.
oc get bc and /health.
managed-agent=true, build=true, language/framework annotations, and optional DevSpaces factory link.
managed-devworkspace=true. agent-applier creates a started DevWorkspace — no push-to-git.
managed-ai-service=true. agent-applier deploys an HTTP service with GET /mcp/smoke.
curl http://<name>:8080/mcp/smoke calls pods_list_in_namespace (k8s-mcp) and monitorDeployments (openshift-mcp). Hub MCP Chat demos: tool calling journey.
The three Golden Paths
| Template | What it does | Steps |
|---|---|---|
| Deploy Agent | Generates real LangGraph / LangChain.js / LangChain4j source, BuildConfig-compiles, and deploys without git push. | resolve → fetch → catalog:write (build=true) → applier builds + deploys |
| DevSpaces AI Workspace | Creates a started DevWorkspace with Continue→LiteLLM for the selected agent language. | catalog:write (managed-devworkspace=true) → applier creates DevWorkspace |
| AI Service with MCP wiring | Deploys a small HTTP service that smoke-calls Kubernetes + OpenShift MCP (/mcp/smoke). |
catalog:write (managed-ai-service=true) → applier deploys Service |
Verify
# Deploy Agent — watch build then pod
oc get bc/<agent-name>
oc logs -f bc/<agent-name>
oc get deploy/<agent-name>
curl -s http://<agent-name>:8080/health
# DevSpaces — open started workspace (no git push)
oc get dw <workspace-name>
# Open DevSpaces dashboard → Open workspace
# AI Service — MCP smoke
oc run curl-smoke --rm -i --restart=Never --image=registry.access.redhat.com/ubi9/ubi-minimal:latest -- \
curl -s http://<service-name>:8080/mcp/smoke