Local testing with kind

End-to-end loop for running Krypton on your laptop. Tested on macOS (OrbStack and Docker Desktop) and Linux.

Prerequisites

• Docker (or OrbStack)
• kind ≥ 0.24 — brew install kind
• Helm ≥ 3.14 — brew install helm
• kubectl — brew install kubectl
• Node 22 + pnpm 10 (for the UI build)
• Go 1.25 (only required for unit tests)

One-liner on macOS:

brew install kind helm kubectl

One-shot install

make deploy-dev

What that runs (hack/local-up.sh):

1. Creates a krypton-dev kind cluster (or reuses one)
2. Builds the React UI and stages it for go:embed
3. Builds five Docker images: manager, control-plane, gateway, krypton-proxy, mcp-hello — all tagged krypton/<name>:dev
4. Loads them into the kind cluster
5. helm upgrade --installs the chart from deploy/helm/krypton
6. Applies the mcp-hello smoke-test agent

Idempotent — re-run any time after editing code.

LLM local test

To test model serving in the same local loop, run:

make deploy-dev-llm

That runs the normal dev setup, preloads the llama.cpp server image into kind, and applies the Qwen2.5 0.5B Model sample from config/samples/llm/qwen2.5-0.5b.yaml.

The model pod still downloads the GGUF weights from Hugging Face on first start, so this path needs network access and can take a few minutes.

After the script prints the port-forward command, wait for the model Deployment and call it:

kubectl -n krypton-system port-forward svc/krypton-gateway 8080:8080 &
kubectl -n models rollout status deploy/qwen2-0-5b --timeout=15m

curl -s http://localhost:8080/v1/models | jq

curl -s http://localhost:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "qwen2-0-5b",
    "messages": [{"role": "user", "content": "Say hi in one word."}]
  }' | jq

You can use the script directly when you want to override the sample:

DEPLOY_LLM=true \
LLM_SAMPLE=config/samples/llm/tinyllama-1.1b.yaml \
LLM_NAME=tinyllama-1-1b \
./hack/local-up.sh

Manual usage after install

# Forward the gateway and control plane in two terminals.
kubectl -n krypton-system port-forward svc/krypton-gateway 8080:8080 &
kubectl -n krypton-system port-forward svc/krypton-control-plane 8090:8090 &

# Open the UI.
open http://localhost:8090/ui/

# Invoke the mcp-hello agent — first call may take a few seconds while
# the pod becomes ready; subsequent calls are sub-millisecond.
curl -X POST http://localhost:8080/v1/agents/agents/mcp-hello/ \
     -H 'Content-Type: application/json' \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Expected response:

{
  "delay": "100ms",
  "echo": "{\"hello\":\"world\"}",
  "method": "POST",
  "path": "/",
  "servedBy": "echo-xxxxxxxx-yyyyy"
}

Smoke test

After port-forwards are running:

make e2e-local

This asserts:

1. A cold invocation returns 200 with the expected echo body
2. A second invocation is warm (sub-3s)
3. The control plane reports phase=Ready

Watching things happen

kubectl get pods -A -w | grep -E 'agents|krypton'
kubectl get agents -A -w

The echo agent runs in always-on mode with minReplicas: 1, so one pod stays up for the lifetime of the CR. The scaler bumps replicas up under load (per spec.concurrency) and back down to minReplicas when load drops.

Teardown

kind delete cluster --name krypton-dev

Troubleshooting

See Troubleshooting & FAQ.