E2E Testing

E2E tests verify that every root example can be built from published tarballs and deployed to real cloud infrastructure. The build phase runs in Docker (same image as the smoke tests) to guarantee a clean environment, but deployment runs directly on the host using chant skills.

Architecture

Docker (build only)                          Host (deploy)
───────────────────                          ─────────────
Dockerfile.smoke-e2e                         Agent + skill
  ├─ Stage 1: bun builds tarballs            cd examples/<name>
  └─ Stage 2: npm install from .tgz          npm run deploy
              npm run build                  npm run teardown

Docker packs @intentius/chant and each lexicon into .tgz tarballs, then installs and builds each example from those tarballs — proving the published packages work in a clean environment.

Host deployment is driven by an AI agent using chant’s built-in skills (e.g. chant-eks, chant-k8s). Each example README describes the deploy flow. The agent runs npm run deploy (and teardown) directly on the host with real cloud credentials.

Examples

Example	Lexicons	Skill	What it deploys
k8s-eks-microservice	AWS + K8s	chant-eks	EKS cluster, K8s workloads, ALB, Route53
k8s-aks-microservice	Azure + K8s	chant-k8s	AKS cluster, K8s workloads, AGIC, Azure DNS
k8s-gke-microservice	GCP + K8s	chant-k8s	GKE cluster, K8s workloads, GCE Ingress, Cloud DNS
gitlab-aws-alb-infra	AWS + GitLab	chant-gitlab-aws-alb-infra	Shared ALB + VPC + ECS cluster via CloudFormation
gitlab-aws-alb-api	AWS + GitLab	chant-gitlab-aws-alb-api	Fargate API service behind ALB
gitlab-aws-alb-ui	AWS + GitLab	chant-gitlab-aws-alb-ui	Fargate UI service behind ALB
flyway-postgresql-gitlab-aws-rds	AWS + Flyway + GitLab	chant-flyway-gitlab-aws-rds	RDS instance + Flyway migration via GitLab CI

Running an E2E test

1. Build in Docker — the smoke test builds tarballs and verifies each example installs and compiles:

   ./test/smoke.sh        # builds all examples from tarballs in Docker

2. Deploy on host — ask your agent to deploy, following the example README:

   Deploy the k8s-eks-microservice example to AWS. My domain is myapp.example.com.

   The agent invokes the appropriate skill and runs the example’s npm scripts directly on the host.

3. Teardown — ask your agent to tear down, or run the example’s teardown script:

   Tear down the k8s-eks-microservice stack.

Required tools

The host needs cloud CLIs and credentials for the examples you’re deploying:

Example group	Host requirements
EKS	AWS CLI, kubectl, jq, AWS credentials with EKS permissions
AKS	Azure CLI, kubectl, Azure credentials with AKS permissions
GKE	gcloud CLI, kubectl, GCP credentials with GKE permissions
GitLab + AWS	AWS CLI, AWS credentials, GITLAB_TOKEN + GITLAB_GROUP

Cleanup

Each example provides its own teardown script (npm run teardown or equivalent). The skill guides the agent through cleanup in the correct order — e.g. for EKS, K8s resources are deleted before the CloudFormation stack to allow the ALB controller to clean up load balancers.

Files

File	Purpose
test/smoke.sh	Builds Docker image, installs and compiles examples from tarballs
test/Dockerfile.smoke-e2e	Two-stage Dockerfile: bun builder → node runner
examples/<name>/README.md	Deploy instructions for each example
examples/<name>/package.json	npm scripts for build, deploy, teardown

Adding a new E2E example

1. Create a root example in examples/<name>/ with working npm run build and npm run deploy scripts
2. Add a corresponding skill in the relevant lexicon (see Skills)
3. Write a README that describes the full deploy flow for the agent
4. Ensure the Dockerfile builds and installs the example from tarballs