Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
#!/usr/bin/env bash
# Run Codex subscription Control UI Playwright guardians against a branch-owned
# Minikube profile with profile-scoped port-forwards and service URLs.
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
ROOT="$(cd "${HERE}/../.." && pwd)"
# shellcheck source=scripts/e2e/load-dotenv.sh
source "${HERE}/load-dotenv.sh"
# shellcheck source=scripts/e2e/admin-credentials.sh
source "${HERE}/admin-credentials.sh"
PROFILE="${MINIKUBE_PROFILE:-${KUBECONTEXT:-}}"
PF_LOG="${E2E_PF_LOG:-/tmp/${PROFILE:-clerum-test}-codex-subscription-playwright-pf.log}"
PF_PID=""
PLAYWRIGHT_DIR="${ROOT}/tests/e2e/playwright"
log() {
printf '[codex-subscription-playwright] %s\n' "$*"
}
die() {
printf '[codex-subscription-playwright] ERROR: %s\n' "$*" >&2
exit 1
}
cleanup() {
if [[ -n "${PF_PID}" ]]; then
kill "${PF_PID}" 2>/dev/null || true
for _ in $(seq 1 20); do
if ! kill -0 "${PF_PID}" 2>/dev/null; then
wait "${PF_PID}" 2>/dev/null || true
return 0
fi
sleep 0.25
done
kill -9 "${PF_PID}" 2>/dev/null || true
wait "${PF_PID}" 2>/dev/null || true
fi
}
wait_http() {
local name="$1"
local url="$2"
local timeout="$3"
local deadline=$((SECONDS + timeout))
while ((SECONDS < deadline)); do
if curl -sf -m 3 "${url}" >/dev/null 2>&1; then
log "${name} ready at ${url}"
return 0
fi
sleep 1
done
die "${name} never became ready at ${url} (see ${PF_LOG})"
}
load_branch_profile_ports() {
local ctx="${KUBECONTEXT:-${MINIKUBE_PROFILE:-}}"
local ports_env="${CLERUM_PROFILE_PORTS_ENV:-${E2E_PROFILE_PORTS_ENV:-}}"
local urls_preconfigured=false
if [[ -n "${CONTROL_UI_URL:-${CONTROL_UI_BASE_URL:-}}" && -n "${CONTROL_API_URL:-${CONTROL_API_BASE_URL:-}}" ]]; then
urls_preconfigured=true
fi
if [[ "${ctx}" =~ ^clerum-(codex|detached)- ]] || [[ "${ctx}" =~ ^clerum-.+-[0-9a-f]{7,8}$ ]]; then
if [[ -z "${ports_env}" ]]; then
ports_env="${HOME}/.cache/clerum/minikube-profiles/${ctx}/ports.env"
fi
if [[ -z "${ports_env}" && "${urls_preconfigured}" != "true" ]]; then
die "branch-owned context ${ctx} requires ports.env from branch-profile preflight, or explicit CONTROL_UI_URL and CONTROL_API_URL"
fi
elif [[ -z "${ports_env}" && -n "${ctx}" ]]; then
ports_env="${HOME}/.cache/clerum/minikube-profiles/${ctx}/ports.env"
fi
if [[ -n "${ports_env}" && -f "${ports_env}" ]]; then
dotenv_load_file "${ports_env}"
fi
if [[ -n "${ports_env}" ]]; then
export CLERUM_PROFILE_PORTS_ENV="${ports_env}"
fi
}
require_random_local_port_for_branch_context() {
local name="$1"
local url="$2"
local default_port="$3"
if [[ ! "${PROFILE}" =~ ^clerum-(codex|detached)- ]] && [[ ! "${PROFILE}" =~ ^clerum-.+-[0-9a-f]{7,8}$ ]]; then
return 0
fi
if [[ "${url}" =~ ^https?://(localhost|127\.0\.0\.1):${default_port}(/|$) ]]; then
die "${PROFILE} must use profile-owned random port-forwards; ${name}=${url} uses shared default port ${default_port}"
fi
}
[[ -n "${PROFILE}" ]] || die "KUBECONTEXT or MINIKUBE_PROFILE is required"
if ! minikube -p "${PROFILE}" status >/dev/null 2>&1; then
die "minikube profile '${PROFILE}' is not running"
fi
if ! kubectl --context="${PROFILE}" cluster-info >/dev/null 2>&1; then
die "kubectl cannot reach context '${PROFILE}'"
fi
[[ -x "${PLAYWRIGHT_DIR}/node_modules/.bin/playwright" ]] || \
die "missing ${PLAYWRIGHT_DIR}/node_modules/.bin/playwright; run npm ci in tests/e2e/playwright"
export MINIKUBE_PROFILE="${PROFILE}"
export KUBECONTEXT="${PROFILE}"
export KUBE_CONTEXT="${PROFILE}"
export KUBECTL_CONTEXT="${PROFILE}"
load_branch_profile_ports
RESOLVED_ADMIN_PASSWORD="$(e2e_resolve_admin_password "${ROOT}" "$(printf '%s%s' 'changeme123' '!')" || true)"
if [[ -z "${RESOLVED_ADMIN_PASSWORD}" ]]; then
die "no admin password is configured in the canonical root .env or process environment"
fi
export TEST_ADMIN_PASSWORD="${TEST_ADMIN_PASSWORD:-${RESOLVED_ADMIN_PASSWORD}}"
export TEST_ADMIN_USERNAME="${TEST_ADMIN_USERNAME:-admin}"
export CONTROL_UI_URL="${CONTROL_UI_URL:-${CONTROL_UI_BASE_URL:-http://127.0.0.1:${CONTROL_UI_PORT:-3000}}}"
export CONTROL_UI_BASE_URL="${CONTROL_UI_BASE_URL:-${CONTROL_UI_URL}}"
export CONTROL_API_URL="${CONTROL_API_URL:-${CONTROL_API_BASE_URL:-http://127.0.0.1:${CONTROL_API_PORT:-8090}}}"
export CONTROL_API_BASE_URL="${CONTROL_API_BASE_URL:-${CONTROL_API_URL}}"
require_random_local_port_for_branch_context "CONTROL_UI_URL" "${CONTROL_UI_URL}" "3000"
require_random_local_port_for_branch_context "CONTROL_API_URL" "${CONTROL_API_URL}" "8090"
log "Syncing Codex subscription Control UI origin into control-api"
sync_args=(--context "${PROFILE}")
if [[ -n "${CLERUM_PROFILE_PORTS_ENV:-}" ]]; then
sync_args+=(--ports-env "${CLERUM_PROFILE_PORTS_ENV}")
fi
"${ROOT}/scripts/minikube/sync-codex-subscription-control-ui-url.sh" "${sync_args[@]}" \
|| die "failed to sync Codex subscription Control UI URL"
if [[ "${E2E_USE_EXISTING_PORT_FORWARDS:-false}" != "true" ]]; then
log "Starting held port-forwards (log: ${PF_LOG})"
"${ROOT}/scripts/minikube/pf-all-stack.sh" --hold >"${PF_LOG}" 2>&1 &
PF_PID="$!"
trap cleanup EXIT INT TERM
else
log "Using existing port-forwards (E2E_USE_EXISTING_PORT_FORWARDS=true)"
fi
wait_http control-ui "${CONTROL_UI_URL%/}/" 90
wait_http control-api "${CONTROL_API_URL%/}/health" 60
# Fail-closed deployment preconditions for the catalog re-sync guardian.
#
# control-api collapses "proxy admin URL unset", "proxy unreachable" and "the
# vendor answered badly" into one identical response — 503, catalog_sync_failed,
# outcome `unavailable` — so no assertion in the spec can separate them. A lane
# missing either precondition would render an error banner and
# codex-subscription-catalog-resync.spec.ts would pass its 503 branch while no
# operator could ever sync a catalog. Checked here, before a browser starts,
# where the diagnosis is unambiguous and costs two API reads.
log "Verifying Codex catalog sync deployment preconditions"
proxy_admin_url="$(kubectl --context="${PROFILE}" -n control-plane get configmap control-api-config \
-o jsonpath='{.data.CODEX_LLM_PROXY_ADMIN_URL}')" \
|| die "cannot read configmap/control-api-config in namespace control-plane"
[[ -n "${proxy_admin_url}" ]] \
|| die "CODEX_LLM_PROXY_ADMIN_URL is empty in configmap/control-api-config; control-api would answer every catalog sync with 503 unavailable without ever reaching the proxy"
proxy_ready="$(kubectl --context="${PROFILE}" -n control-plane get deployment codex-llm-proxy \
-o jsonpath='{.status.readyReplicas}')" \
|| die "cannot read deployment/codex-llm-proxy in namespace control-plane"
[[ "${proxy_ready:-0}" -ge 1 ]] \
|| die "deployment/codex-llm-proxy has no ready replica (readyReplicas='${proxy_ready:-0}'); control-api would answer every catalog sync with 503 unavailable"
log "Preconditions OK (proxy admin URL set, codex-llm-proxy readyReplicas=${proxy_ready})"
log "Running Control UI Playwright guardians (CONTROL_UI_URL=${CONTROL_UI_URL})"
(
cd "${PLAYWRIGHT_DIR}"
PLAYWRIGHT_BROWSERS_PATH="${PLAYWRIGHT_DIR}/.playwright-browsers" \
./node_modules/.bin/playwright test \
--config playwright.codex-subscription.config.ts \
--project=control-ui
)