feat(ci): add PR merge and deploy lifecycle

.github/workflows/ci.yml

@@ -9,24 +9,19 @@ on:
 permissions:
   contents: read
 
+concurrency:
+  group: ci-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
 jobs:
-  test:
-    runs-on: ubuntu-latest
+  quality:
+    runs-on: [self-hosted, linux, x64, platform-ci, agent-trajectory-profiler]
+    timeout-minutes: 20
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-node@v4
+      - name: Check out source
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
         with:
-          node-version: 20
-      - name: Fetch pinned public dependencies
-        run: |
-          git clone \
-            https://github.com/opsle/context-firewall.git \
-            ../context-firewall
-          git -C ../context-firewall checkout \
-            953c48f1cfd154d6b7ed10b51b87fe54e4df45f2
-          git clone \
-            https://github.com/opsle/decision-evidence-protocol.git \
-            ../decision-evidence-protocol
-          git -C ../decision-evidence-protocol checkout \
-            b17ae3b41cea7cb0b9e0befe43e885b5aa0e4a09
-      - run: npm run verify
+          fetch-depth: 0
+
+      - name: TEST
+        run: ops/ci/test

.github/workflows/deploy.yml

@@ -0,0 +1,49 @@
+name: Deploy
+
+# DEPLOY and VERIFY run only for the exact SHA that the main-branch CI run just
+# tested successfully. workflow_run always executes this file as it exists on the
+# default branch, so a pull request cannot change what deploys. The deploy runner
+# additionally refuses every job that is not this file on main.
+on:
+  workflow_run:
+    workflows: [CI]
+    types: [completed]
+    branches: [main]
+
+permissions:
+  contents: read
+
+concurrency:
+  group: deploy-agent-trajectory-profiler
+  cancel-in-progress: false
+
+jobs:
+  deploy:
+    # AGENT_TRAJECTORY_PROFILER_DEPLOY_ENABLED is the single switch that makes GitHub the release owner.
+    # Leave it unset while another release path is authoritative.
+    if: >-
+      vars.AGENT_TRAJECTORY_PROFILER_DEPLOY_ENABLED == 'true' &&
+      github.event.workflow_run.conclusion == 'success' &&
+      github.event.workflow_run.event == 'push' &&
+      github.event.workflow_run.head_branch == 'main'
+    runs-on: [self-hosted, linux, x64, agent-trajectory-profiler-deploy]
+    timeout-minutes: 30
+    env:
+      SHA: ${{ github.event.workflow_run.head_sha }}
+    steps:
+      - name: Check out the tested SHA
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
+        with:
+          ref: ${{ github.event.workflow_run.head_sha }}
+
+      - name: DEPLOY
+        id: deploy
+        run: |
+          status=0
+          ops/ci/deploy "$SHA" || status=$?
+          if [[ $status -eq 75 ]]; then echo "superseded=true" >>"$GITHUB_OUTPUT"; exit 0; fi
+          exit "$status"
+
+      - name: VERIFY
+        if: steps.deploy.outputs.superseded != 'true'
+        run: ops/ci/verify "$SHA"

.gitignore

@@ -2,3 +2,4 @@ node_modules/
 coverage/
 *.log
 .DS_Store
+.deps/

ops/ci/README.md

@@ -0,0 +1,17 @@
+# CI and Deployment Lifecycle
+
+This repository implements the canonical self-hosted deployment lifecycle:
+
+```
+AI -> PR -> TEST -> MERGE -> TEST merged SHA -> DEPLOY -> VERIFY
+```
+
+## Structure
+- `ops/ci/test`: Runs test suite and static checks
+- `ops/ci/deploy <SHA>`: Safely deploys tested release to `/opt/opsle-components/agent-trajectory-profiler`
+- `ops/ci/verify <SHA>`: Verifies receipt and tests deployed installation
+- `ops/ci/operator-gate`: Pre-flight operator environment validation
+- `ops/ci/install-test-runner`: Sets up self-hosted CI runner
+- `ops/ci/install-deploy-runner`: Sets up self-hosted deploy runner
+- `.github/workflows/ci.yml`: Runs `ops/ci/test` on pull requests and main pushes
+- `.github/workflows/deploy.yml`: Deploys only verified merge SHAs on main

ops/ci/deploy

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Fail-closed deployment for agent-trajectory-profiler
+# Usage: ops/ci/deploy <SHA>
+
+sha="${1:-}"
+if [[ -z "$sha" ]]; then
+  echo "usage: $0 <SHA>" >&2
+  exit 2
+fi
+
+root="${OPSLE_COMPONENT_ROOT:-/opt/opsle-components/agent-trajectory-profiler}"
+releases_dir="$root/releases"
+current_symlink="$root/current"
+receipts_dir="$root/receipts"
+
+# Reject if workspace is dirty
+if [[ -n "$(git status --porcelain)" ]]; then
+  echo "deploy: the checkout is not clean" >&2
+  exit 1
+fi
+
+# Reject if sha is superseded on origin/main
+current_origin_sha="$(git rev-parse origin/main)"
+if [[ "$sha" != "$current_origin_sha" ]]; then
+  echo "deploy: $sha is not the tip of main ($current_origin_sha); superseded, nothing deployed" >&2
+  exit 75
+fi
+
+target_dir="$releases_dir/$sha"
+mkdir -p "$target_dir" "$receipts_dir"
+
+# Export git archive to release directory
+git archive "$sha" | tar -x -C "$target_dir"
+
+# Atomically flip symlink
+ln -sfn "$target_dir" "$current_symlink.tmp"
+mv -Tf "$current_symlink.tmp" "$current_symlink"
+
+# Write deployment receipt
+receipt_file="$receipts_dir/deploy-$sha.json"
+cat >"$receipt_file" <<EOF
+{
+  "component": "agent-trajectory-profiler",
+  "sha": "$sha",
+  "deployed_at": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
+  "deployer": "${USER:-unknown}",
+  "status": "deployed",
+  "path": "$target_dir"
+}
+EOF
+
+echo "deploy: successfully deployed $sha to $target_dir"

ops/ci/install-deploy-runner

@@ -0,0 +1,92 @@
+#!/usr/bin/env bash
+# One-time OPERATOR gate (run as root on the VPS host): install the deploy
+# runner used only by .github/workflows/deploy.yml.
+#   sudo bash ops/ci/install-deploy-runner < registration-token-file
+set -Eeuo pipefail
+
+readonly account=agent-trajectory-deployer
+readonly home=/var/lib/agent-trajectory-deployer
+readonly runner_dir="$home/runner"
+readonly support=/usr/local/lib/agent-trajectory-deploy-runner
+readonly release_root=/opt/opsle-components/agent-trajectory-profiler
+readonly repo=opsle/agent-trajectory-profiler
+readonly repo_url="https://github.com/$repo"
+readonly unit=agent-trajectory-deploy-runner.service
+readonly version=2.336.0
+readonly archive_sha=04cf0be1aff4c3ec3554466c39124ca250e3effd8873bb7e8d68535aa9505d5d
+readonly archive_url="https://github.com/actions/runner/releases/download/v$version/actions-runner-linux-x64-$version.tar.gz"
+
+[[ $EUID -eq 0 ]] || { echo "Run as root." >&2; exit 77; }
+command -v setfacl >/dev/null || { echo "setfacl is required." >&2; exit 69; }
+IFS= read -r token
+[[ $token =~ ^[A-Za-z0-9]{20,64}$ ]] || { echo "Expected one registration token on stdin." >&2; exit 64; }
+
+mkdir -p "$release_root"
+id "$account" >/dev/null 2>&1 ||
+  useradd --system --home-dir "$home" --create-home --shell /usr/sbin/nologin "$account"
+install -d -o "$account" -g "$account" -m 0700 "$runner_dir"
+
+setfacl -R -m "u:$account:rwX" "$release_root" 2>/dev/null || true
+setfacl -R -d -m "u:$account:rwX" "$release_root" 2>/dev/null || true
+setfacl -R -m "u:deploy:rwX" "$release_root" 2>/dev/null || true
+setfacl -R -d -m "u:deploy:rwX" "$release_root" 2>/dev/null || true
+
+install -d -o root -g root -m 0755 "$support"
+cat >"$support/job-started.sh" <<'HOOK'
+#!/usr/bin/env bash
+set -Eeuo pipefail
+[[ ${GITHUB_WORKFLOW_REF:-} == "opsle/agent-trajectory-profiler/.github/workflows/deploy.yml@refs/heads/main" ]] &&
+  [[ ${GITHUB_EVENT_NAME:-} == workflow_run ]] || {
+  echo "agent-trajectory-profiler deploy runner refuses ${GITHUB_WORKFLOW_REF:-unknown} (${GITHUB_EVENT_NAME:-unknown})" >&2
+  exit 78
+}
+HOOK
+cat >"$support/job-completed.sh" <<'HOOK'
+#!/usr/bin/env bash
+set -Eeuo pipefail
+find /var/lib/agent-trajectory-deployer/runner/_work -xdev -mindepth 1 -delete
+HOOK
+chmod 0755 "$support/job-started.sh" "$support/job-completed.sh"
+
+archive="$(mktemp)"
+trap 'rm -f -- "$archive"' EXIT
+curl --fail --silent --show-error --location --output "$archive" "$archive_url"
+[[ "$(sha256sum "$archive" | cut -d' ' -f1)" == "$archive_sha" ]] || { echo "Runner archive hash mismatch." >&2; exit 65; }
+tar -xzf "$archive" -C "$runner_dir"
+chown -R "$account:$account" "$runner_dir"
+
+runuser -u "$account" -- env HOME="$home" "$runner_dir/config.sh" \
+  --unattended --replace \
+  --url "$repo_url" --token "$token" \
+  --name agent-trajectory-deploy-vps --labels agent-trajectory-profiler-deploy --work "$runner_dir/_work"
+
+cat >"/etc/systemd/system/$unit" <<UNIT
+[Unit]
+Description=Agent Trajectory Profiler deploy runner (deploy.yml on main only)
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+User=$account
+Group=$account
+WorkingDirectory=$runner_dir
+ExecStart=$runner_dir/bin/Runner.Listener run --once
+Restart=always
+RestartSec=10s
+Environment=HOME=$home
+Environment=ACTIONS_RUNNER_HOOK_JOB_STARTED=$support/job-started.sh
+Environment=ACTIONS_RUNNER_HOOK_JOB_COMPLETED=$support/job-completed.sh
+NoNewPrivileges=true
+PrivateTmp=true
+ProtectSystem=strict
+ProtectHome=true
+ReadWritePaths=$home /opt/opsle-components/agent-trajectory-profiler
+InaccessiblePaths=-/home/deploy -/root -/var/run/docker.sock
+UMask=0077
+
+[Install]
+WantedBy=multi-user.target
+UNIT
+systemctl daemon-reload
+systemctl enable --now "$unit"
+echo "Installed $unit (label agent-trajectory-profiler-deploy)."

ops/ci/install-test-runner

@@ -0,0 +1,90 @@
+#!/usr/bin/env bash
+# One-time OPERATOR gate (run as root on the VPS host): install the repo-scoped
+# TEST runner used by .github/workflows/ci.yml.
+#   sudo bash ops/ci/install-test-runner < registration-token-file
+set -Eeuo pipefail
+
+readonly account=agent-trajectory-profiler-ci
+readonly home=/var/lib/agent-trajectory-profiler-ci
+readonly runner_dir="$home/runner"
+readonly support=/usr/local/lib/agent-trajectory-profiler-ci-runner
+readonly repo_url=https://github.com/opsle/agent-trajectory-profiler
+readonly unit=agent-trajectory-profiler-ci-runner.service
+readonly version=2.336.0
+readonly archive_sha=04cf0be1aff4c3ec3554466c39124ca250e3effd8873bb7e8d68535aa9505d5d
+readonly archive_url="https://github.com/actions/runner/releases/download/v$version/actions-runner-linux-x64-$version.tar.gz"
+
+[[ $EUID -eq 0 ]] || { echo "Run as root." >&2; exit 77; }
+IFS= read -r token
+[[ $token =~ ^[A-Za-z0-9]{20,64}$ ]] || { echo "Expected one registration token on stdin." >&2; exit 64; }
+command -v git >/dev/null || { echo "git is required on the host." >&2; exit 69; }
+
+id "$account" >/dev/null 2>&1 ||
+  useradd --system --home-dir "$home" --create-home --shell /usr/sbin/nologin "$account"
+install -d -o "$account" -g "$account" -m 0700 "$runner_dir" "$home/cache"
+
+install -d -o root -g root -m 0755 "$support"
+cat >"$support/job-started.sh" <<'HOOK'
+#!/usr/bin/env bash
+set -Eeuo pipefail
+for variable in DATABASE_URL AGENT_TRAJECTORY_PROFILER_DEPLOY_SSH_KEY OPENAI_API_KEY OPSLE_CODEX_AUTH; do
+  [[ -z ${!variable:-} ]] || { echo "TEST runner rejected $variable" >&2; exit 78; }
+done
+for path in /home/deploy/.config/gh/hosts.yml /var/run/docker.sock; do
+  if [[ -r $path || -w $path ]]; then echo "TEST runner can reach $path" >&2; exit 78; fi
+done
+HOOK
+cat >"$support/job-completed.sh" <<'HOOK'
+#!/usr/bin/env bash
+set -Eeuo pipefail
+find /var/lib/agent-trajectory-profiler-ci/runner/_work -xdev -mindepth 1 -delete
+HOOK
+chmod 0755 "$support/job-started.sh" "$support/job-completed.sh"
+
+archive="$(mktemp)"
+trap 'rm -f -- "$archive"' EXIT
+curl --fail --silent --show-error --location --output "$archive" "$archive_url"
+[[ "$(sha256sum "$archive" | cut -d' ' -f1)" == "$archive_sha" ]] || { echo "Runner archive hash mismatch." >&2; exit 65; }
+tar -xzf "$archive" -C "$runner_dir"
+chown -R "$account:$account" "$runner_dir"
+
+runuser -u "$account" -- env HOME="$home" "$runner_dir/config.sh" \
+  --unattended --replace \
+  --url "$repo_url" --token "$token" \
+  --name platform-ci-agent-trajectory-profiler-vps --labels platform-ci,agent-trajectory-profiler --work "$runner_dir/_work"
+
+cat >"/etc/systemd/system/$unit" <<UNIT
+[Unit]
+Description=Agent Trajectory Profiler TEST runner (ci.yml)
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+User=$account
+Group=$account
+WorkingDirectory=$runner_dir
+ExecStart=$runner_dir/bin/Runner.Listener run --once
+Restart=always
+RestartSec=10s
+Environment=HOME=$home
+Environment=CI=true
+Environment=RUNNER_TOOL_CACHE=$home/cache/tool
+Environment=ACTIONS_RUNNER_HOOK_JOB_STARTED=$support/job-started.sh
+Environment=ACTIONS_RUNNER_HOOK_JOB_COMPLETED=$support/job-completed.sh
+NoNewPrivileges=true
+PrivateTmp=true
+PrivateDevices=true
+ProtectSystem=strict
+ProtectHome=true
+ReadWritePaths=$home
+InaccessiblePaths=-/home/deploy -/root -/var/run/docker.sock
+CapabilityBoundingSet=
+RestrictSUIDSGID=true
+UMask=0077
+
+[Install]
+WantedBy=multi-user.target
+UNIT
+systemctl daemon-reload
+systemctl enable --now "$unit"
+echo "Installed $unit (labels platform-ci,agent-trajectory-profiler)."

ops/ci/operator-gate

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+# Operator helper for the root-only gates for opsle/agent-trajectory-profiler.
+#   bash ops/ci/operator-gate test        install the TEST runner
+#   bash ops/ci/operator-gate deploy      install the deploy runner
+set -Eeuo pipefail
+cd "$(dirname "${BASH_SOURCE[0]}")/../.."
+
+readonly repo=opsle/agent-trajectory-profiler
+
+token() { gh api -X POST "repos/$repo/actions/runners/registration-token" --jq .token; }
+online() {
+  local _ found
+  for _ in {1..30}; do
+    found="$(gh api "repos/$repo/actions/runners" --jq ".runners[] | select(.name==\"$1\" and .status==\"online\") | .name")"
+    if [[ -n $found ]]; then echo "runner $1 is online"; return 0; fi
+    sleep 2
+  done
+  echo "runner $1 did not come online" >&2; return 1
+}
+
+[[ $# -gt 0 ]] || { echo "usage: ops/ci/operator-gate test|deploy ..." >&2; exit 64; }
+sudo -v
+for step in "$@"; do
+  case "$step" in
+    test)
+      token | sudo bash ops/ci/install-test-runner
+      online platform-ci-agent-trajectory-profiler-vps ;;
+    deploy)
+      token | sudo bash ops/ci/install-deploy-runner
+      online agent-trajectory-deploy-vps ;;
+    *) echo "unknown step: $step" >&2; exit 64 ;;
+  esac
+done