Kaytoo e2e (kind)

Local stack: Mermin + OpenSearch (k8s/mermin-stack-values.yaml, embedded in k8s/bootstrap.yaml) and Kaytoo (../helm/kaytoo) with harness overrides in k8s/kaytoo-values.yaml.

Commands (repo root)

Command	Purpose
npm run e2e:up	kind cluster, image build/load, bootstrap Job, port-forwards
npm run e2e:down	Stop forwards, delete cluster, remove image + e2e/.generated/
npm run e2e:verify	Smoke: rollout, /health, /chat, OpenSearch-backed checks
npm run e2e:status	URLs + quick pod list
npm run e2e:logs	Tail Kaytoo in-cluster
npm run e2e:dev	Run Kaytoo on the host against forwarded OpenSearch

Host Helm is not required - the bootstrap Job runs Helm in-cluster.

Stack

Cluster	kind name kaytoo-e2e
Namespace	elastiflow
Kaytoo	Helm release; HTTP chat svc/kaytoo-chat:8080 (forwarded to 18080)
Image	kaytoo-e2e:local from repo Dockerfile
LLM	Real API; credentials from .env via generated Helm fragment (below)

Configuration

Source of truth: repo-root .env (see .env.example). Optional: E2E_ENV_FILE.

Variable	e2e:up	Notes
OPENSEARCH_PASSWORD	required	Mermin/OpenSearch admin; also baked into rendered Kaytoo values
LLM_BASE_URL	required	OpenAI-compatible URL
LLM_API_KEY	required
LLM_MODEL	optional

For e2e:verify / e2e:dev against host forwards, set e.g. OPENSEARCH_URL=https://127.0.0.1:9200, OPENSEARCH_USERNAME=admin, OPENSEARCH_TLS_INSECURE=true, OPENSEARCH_INDEX_PATTERN=elastiflow-flow-codex-*.

Interactive e2e:up: if LLM_* is missing, the script prompts once and appends to .env (or E2E_ENV_FILE when used).

e2e/.generated/ (gitignored)

Created during e2e:up (and by npm run e2e:llm-overlay for the LLM file only). Do not commit.

Artifact	Role
kubeconfig-kind-kaytoo-e2e	Host / kind CLI - API server https://127.0.0.1:...
kubeconfig-kind-kaytoo-e2e.docker	Docker kubectl port-forward sidecars - same cluster, API host host.docker.internal
kaytoo-values.rendered.yaml	k8s/kaytoo-values.yaml + real OPENSEARCH_PASSWORD -> copied to node as kaytoo-values.yaml (secrets)
values-e2e.llm.local.json	LLM Helm -f from .env via e2e/write-e2e-llm-overlay.mjs

Legacy e2e/.kubeconfig-kind-* at the e2e/ root is moved into .generated/ on first use.

Bring up

npm run e2e:up

1. kind cluster kaytoo-e2e (prompts if it already exists).
2. docker build -> kaytoo-e2e:local -> kind load.
3. Writes .generated/ (LLM JSON, rendered Kaytoo values), copies chart + those files to the control-plane node (/kaytoo-chart/), applies bootstrap Job.
4. Job: metrics-server, Mermin stack, helm upgrade -i kaytoo ... -f kaytoo-values.yaml -f values-e2e.llm.local.json.
5. Docker-based port-forwards: OpenSearch 9200, Dashboards 5601, chat 18080 -> 8080.

URLs after up: OpenSearch https://127.0.0.1:9200, Dashboards http://localhost:5601, chat http://127.0.0.1:18080.

kubectl (from repo root):

export KUBECONFIG="$(pwd)/e2e/.generated/kubeconfig-kind-kaytoo-e2e"
kubectl -n elastiflow get pods
kubectl -n elastiflow logs deploy/kaytoo -c kaytoo -f

Verify

With forwards up and OpenSearch reachable on the host:

export KUBECONFIG="$(pwd)/e2e/.generated/kubeconfig-kind-kaytoo-e2e"
npm run e2e:verify

verify starts a temporary chat port-forward if needed.

Host Kaytoo (optional)

After e2e:up:

npm run e2e:dev

Uses the same .env; point OpenSearch at the forwards (e.g. OPENSEARCH_URL=https://127.0.0.1:9200).

Tear down

npm run e2e:down

Removes forwards, cluster, kaytoo-e2e:local image, e2e/.generated/, and stale e2e/.kubeconfig-kind-*.

Manifests to keep in sync

What	Where
Mermin / OpenSearch	k8s/mermin-stack-values.yaml + embedded key mermin-stack-values.yaml in k8s/bootstrap.yaml
Kaytoo chart	../helm/kaytoo
Kaytoo e2e overrides	k8s/kaytoo-values.yaml (placeholders only; secrets only in .generated/ at deploy)

Helm preview (no secrets in base -f):

helm template kaytoo ./helm/kaytoo -n elastiflow -f ./e2e/k8s/kaytoo-values.yaml

With LLM overlay (after set -a && source .env && set +a, run npm run e2e:llm-overlay if the JSON is not already there):

helm template kaytoo ./helm/kaytoo -n elastiflow \
  -f ./e2e/k8s/kaytoo-values.yaml \
  -f ./e2e/.generated/values-e2e.llm.local.json

To regenerate only the LLM fragment: set -a && source .env && set +a then npm run e2e:llm-overlay.