PaperClipAI/.github/workflows/docker-runner-check.yml

name: Docker Runner check

on:
  workflow_dispatch:
    inputs:
      target_branch:
        description: "Branch in this repository to build; resolved to one immutable commit before checkout"
        type: string
        required: false
      publish_eval_image:
        description: "Publish an immutable Daytona qualification image on the EC2 fleet (no provider credentials)"
        type: boolean
        default: false
      verify_source:
        description: "Run broad source checks on EC2"
        type: boolean
        default: false
      verify_runner:
        description: "Run the complete offline Runner verify command on GitHub-hosted Ubuntu without building an image"
        type: boolean
        default: false
      diagnose_ajv_pack:
        description: "Diagnose only pinned AJV npm directory packing on Linux (no Runner build/tests)"
        type: boolean
        default: false
      expected_source_sha:
        description: "Require this immutable target commit (required for verify_runner)"
        type: string
        required: false
      expected_resolved_lock_sha256:
        description: "Require this reviewed resolved dependency lock (required for verify_runner)"
        type: string
        required: false
      verify_public_install:
        description: "Build and verify clean public npm installation on EC2 (no provider credentials)"
        type: boolean
        default: false
      build_eval_viewer:
        description: "Build the canonical eval report viewer on EC2"
        type: boolean
        default: false
  pull_request:
    paths:
      - .github/workflows/docker-runner-check.yml
      - Dockerfile
      - .dockerignore
      - scripts/check-docker-runner-cache.sh
      - packages/paperclip-runner/rust-toolchain.toml
      - packages/paperclip-runner/runner/**
      - packages/paperclip-runner/protocol/**

permissions: {}

concurrency:
  # Publishing a newer image must not discard an earlier verification's evidence.
  group: docker-runner-check-${{ github.event.pull_request.number || github.ref }}-${{ inputs.diagnose_ajv_pack && 'ajv-diagnostic' || inputs.verify_runner && 'runner-verification' || inputs.verify_source && 'verification' || inputs.verify_public_install && 'public-install' || inputs.build_eval_viewer && 'viewer' || 'image' }}
  cancel-in-progress: true

jobs:
  runner:
    if: github.event_name == 'pull_request'
    name: Compile isolated native Runner
    runs-on: ubuntu-latest
    timeout-minutes: 20
    permissions:
      contents: read
    steps:
      - name: Checkout repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          persist-credentials: false
      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@37fe631027851001ddb9b187196cc803df7f5f0e # v4
      # Compile the real target, then change source in a disposable context.
      # A fresh builder must import dependencies and produce changed binary metadata.
      # The baseline build anonymously seeds from the public BuildKit cache at
      # ghcr.io/paperclipai/paperclip:buildcache-{amd64,arm64}; the verification
      # build imports only this run's locally exported cache on a fresh builder.
      # No registry credentials or image publication.
      - name: Verify native build and dependency cache reuse
        run: bash scripts/check-docker-runner-cache.sh


  authorize_manual:
    if: github.event_name == 'workflow_dispatch'
    runs-on: ubuntu-latest
    permissions:
      contents: read
    timeout-minutes: 5
    outputs:
      target_sha: ${{ steps.authorize.outputs.target_sha }}
    steps:
      - name: Authorize an explicit maintainer image build
        id: authorize
        env:
          GH_TOKEN: ${{ github.token }}
          REPOSITORY: ${{ github.repository }}
          REPOSITORY_ID: ${{ github.repository_id }}
          ACTOR_ID: ${{ github.actor_id }}
          TRIGGERING_ACTOR: ${{ github.triggering_actor }}
          ALLOWED_IDS: ${{ vars.AWS_CI_TRUSTED_USER_IDS }}
          TARGET_BRANCH: ${{ inputs.target_branch || github.ref_name }}
          EXPECTED_SOURCE_SHA: ${{ inputs.expected_source_sha }}
          EXPECTED_LOCK_SHA256: ${{ inputs.expected_resolved_lock_sha256 }}
          VERIFY_RUNNER: ${{ inputs.verify_runner }}
          DIAGNOSE_AJV: ${{ inputs.diagnose_ajv_pack }}
          OTHER_MODE: ${{ inputs.publish_eval_image || inputs.verify_source || inputs.verify_public_install || inputs.build_eval_viewer }}
        run: |
          set -euo pipefail
          test "$REPOSITORY" = paperclipai/paperclip
          test "$REPOSITORY_ID" = 1170821064
          jq -e 'type == "array" and length > 0 and all(.[]; type == "number")' <<< "$ALLOWED_IDS" >/dev/null
          triggering_id="$(gh api "users/$TRIGGERING_ACTOR" --jq .id)"
          for id in "$ACTOR_ID" "$triggering_id"; do
            jq -e --argjson id "$id" 'index($id) != null' <<< "$ALLOWED_IDS" >/dev/null
          done
          target_sha="$(gh api "repos/$REPOSITORY/git/ref/heads/$TARGET_BRANCH" --jq '.object.sha')"
          [[ "$target_sha" =~ ^[0-9a-f]{40}$ ]]
          if [ "$VERIFY_RUNNER" = true ]; then
            [[ "$EXPECTED_SOURCE_SHA" =~ ^[0-9a-f]{40}$ ]]
            [[ "$EXPECTED_LOCK_SHA256" =~ ^[0-9a-f]{64}$ ]]
            test "$OTHER_MODE" != true
          fi
          if [ "$DIAGNOSE_AJV" = true ]; then
            test "$VERIFY_RUNNER" != true
            test "$OTHER_MODE" != true
            test "$EXPECTED_SOURCE_SHA" = 34f49a201a804564cfae81e425266b3f9f987e30
            test "$EXPECTED_LOCK_SHA256" = 5ae57d1ddd475691dac1f1cfbd4836e54f7da1956b47e6e705b218ae3070ae2e
          fi
          if [ -n "$EXPECTED_SOURCE_SHA" ]; then
            test "$target_sha" = "$EXPECTED_SOURCE_SHA"
          fi
          echo "target_sha=$target_sha" >> "$GITHUB_OUTPUT"

  manual_runner_verify:
    name: Full offline Runner verification on GitHub-hosted Ubuntu
    if: github.event_name == 'workflow_dispatch' && inputs.verify_runner
    needs: authorize_manual
    runs-on: ubuntu-latest
    timeout-minutes: 55
    permissions:
      contents: read
    env:
      CI: "true"
      PAPERCLIP_TELEMETRY_ENABLED: "false"
      NPM_CONFIG_AUDIT: "false"
      NPM_CONFIG_FUND: "false"
      NPM_CONFIG_UPDATE_NOTIFIER: "false"
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ needs.authorize_manual.outputs.target_sha }}
          persist-credentials: false
      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: 24.21.0
          package-manager-cache: false
      - uses: pnpm/action-setup@0977fd99725f1db4007ccb2928dbb4e90d06cc86 # v6
        with:
          version: 9.15.4
      - name: Record exact source and install dependencies
        timeout-minutes: 8
        env:
          SOURCE_SHA: ${{ needs.authorize_manual.outputs.target_sha }}
          EXPECTED_LOCK_SHA256: ${{ inputs.expected_resolved_lock_sha256 }}
        run: |
          set -euo pipefail
          test "$(git rev-parse HEAD)" = "$SOURCE_SHA"
          mkdir -p remote-runner-verification
          git rev-parse HEAD > remote-runner-verification/source.txt
          cp pnpm-lock.yaml remote-runner-verification/original-pnpm-lock.yaml
          sha256sum pnpm-lock.yaml > remote-runner-verification/original-lock.sha256
          # A stale stacked-branch lock may be resolved only to the reviewed overlay.
          pnpm install --resolution-only --ignore-scripts --no-frozen-lockfile
          cp pnpm-lock.yaml remote-runner-verification/resolved-pnpm-lock.yaml
          sha256sum pnpm-lock.yaml > remote-runner-verification/resolved-lock.sha256
          git diff -- pnpm-lock.yaml > remote-runner-verification/lock.diff
          printf '%s  pnpm-lock.yaml\n' "$EXPECTED_LOCK_SHA256" | sha256sum --check --strict
          pnpm install --frozen-lockfile --ignore-scripts
      - name: Select pinned Rust and install browser dependencies
        timeout-minutes: 8
        run: |
          set -euo pipefail
          cd packages/paperclip-runner
          rustup show
          rustc --version --verbose > ../../remote-runner-verification/rust.txt
          pnpm exec playwright install --with-deps chromium
      - name: Run the complete offline Runner verification
        timeout-minutes: 35
        run: |
          set -uo pipefail
          status=0
          timeout --signal=TERM --kill-after=20s 34m \
            pnpm --filter @paperclipai/paperclip-runner verify \
            > remote-runner-verification/verify.log 2>&1 || status=$?
          printf '%s\n' "$status" > remote-runner-verification/exit-code.txt
          tail -n 100 remote-runner-verification/verify.log
          exit "$status"
      - name: Retain Runner verification evidence
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: runner-full-verification-${{ github.run_id }}
          path: |
            remote-runner-verification/
            packages/paperclip-runner/**/playwright-report/
            packages/paperclip-runner/**/test-results/
          retention-days: 14

  manual_ajv_pack_diagnostic:
    name: Pinned AJV directory-pack diagnostic on Linux
    if: github.event_name == 'workflow_dispatch' && inputs.diagnose_ajv_pack
    needs: authorize_manual
    runs-on: ubuntu-latest
    timeout-minutes: 10
    permissions:
      contents: read
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ needs.authorize_manual.outputs.target_sha }}
          persist-credentials: false
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ github.workflow_sha }}
          path: .ajv-diagnostic-driver
          sparse-checkout: .github/scripts/diagnose-ajv-pack.py
          sparse-checkout-cone-mode: false
          persist-credentials: false
      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: 24.21.0
          package-manager-cache: false
      - uses: pnpm/action-setup@0977fd99725f1db4007ccb2928dbb4e90d06cc86 # v6
        with:
          version: 9.15.4
      - name: Install only exact reviewed dependency graph without scripts
        timeout-minutes: 6
        env:
          EXPECTED_LOCK_SHA256: ${{ inputs.expected_resolved_lock_sha256 }}
        run: |
          set -euo pipefail
          umask 077
          mkdir -p ajv-pack-diagnostic "$RUNNER_TEMP/ajv-install-home"
          export HOME="$RUNNER_TEMP/ajv-install-home"
          export NPM_CONFIG_USERCONFIG="$HOME/user.npmrc"
          export NPM_CONFIG_GLOBALCONFIG="$HOME/global.npmrc"
          touch "$NPM_CONFIG_USERCONFIG" "$NPM_CONFIG_GLOBALCONFIG"
          test "$(git rev-parse HEAD)" = 34f49a201a804564cfae81e425266b3f9f987e30
          test "$(node --version)" = v24.21.0
          test "$(npm --version)" = 11.19.0
          test "$(pnpm --version)" = 9.15.4
          git rev-parse HEAD > ajv-pack-diagnostic/source.txt
          git -C .ajv-diagnostic-driver rev-parse HEAD > ajv-pack-diagnostic/driver-source.txt
          cp pnpm-lock.yaml ajv-pack-diagnostic/original-pnpm-lock.yaml
          pnpm install --resolution-only --ignore-scripts --no-frozen-lockfile > ajv-pack-diagnostic/resolution.log 2>&1
          cp pnpm-lock.yaml ajv-pack-diagnostic/resolved-pnpm-lock.yaml
          printf '%s  pnpm-lock.yaml\n' "$EXPECTED_LOCK_SHA256" | sha256sum --check --strict
          pnpm install --frozen-lockfile --ignore-scripts > ajv-pack-diagnostic/install.log 2>&1
      - name: Compare installed-layout and identical plain-package packing
        timeout-minutes: 3
        run: python3 .ajv-diagnostic-driver/.github/scripts/diagnose-ajv-pack.py
      - name: Retain diagnostic evidence even on npm failure
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: ajv-pack-diagnostic-${{ github.run_id }}
          path: ajv-pack-diagnostic/
          retention-days: 14

  manual_image:
    name: Build and verify on EC2
    if: github.event_name == 'workflow_dispatch' && !inputs.verify_runner && !inputs.diagnose_ajv_pack
    needs: authorize_manual
    runs-on: runs-on/fleet=paperclip-public-pr-x64/env=public-ci
    timeout-minutes: 90
    permissions:
      contents: read
      packages: write
    steps:
      - name: Authorize an explicit maintainer image build
        env:
          GH_TOKEN: ${{ github.token }}
          REPOSITORY: ${{ github.repository }}
          REPOSITORY_ID: ${{ github.repository_id }}
          ACTOR_ID: ${{ github.actor_id }}
          TRIGGERING_ACTOR: ${{ github.triggering_actor }}
          ALLOWED_IDS: ${{ vars.AWS_CI_TRUSTED_USER_IDS }}
        run: |
          set -euo pipefail
          test "$REPOSITORY" = paperclipai/paperclip
          test "$REPOSITORY_ID" = 1170821064
          jq -e 'type == "array" and length > 0 and all(.[]; type == "number")' <<< "$ALLOWED_IDS" >/dev/null
          triggering_id="$(gh api "users/$TRIGGERING_ACTOR" --jq .id)"
          for id in "$ACTOR_ID" "$triggering_id"; do
            jq -e --argjson id "$id" 'index($id) != null' <<< "$ALLOWED_IDS" >/dev/null
          done
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
        with:
          ref: ${{ needs.authorize_manual.outputs.target_sha }}
          persist-credentials: false
      - uses: pnpm/action-setup@0977fd99725f1db4007ccb2928dbb4e90d06cc86 # v6
        with:
          version: 9.15.4
      - uses: actions/setup-node@820762786026740c76f36085b0efc47a31fe5020 # v7
        with:
          node-version: 24
          package-manager-cache: false
      - name: Resolve source dependencies without lifecycle scripts
        run: |
          pnpm install --resolution-only --ignore-scripts --no-frozen-lockfile
          mkdir -p remote-verification
          sha256sum pnpm-lock.yaml > remote-verification/lock.sha256
          git rev-parse HEAD > remote-verification/source.txt
      - uses: docker/setup-buildx-action@37fe631027851001ddb9b187196cc803df7f5f0e # v4
      - if: inputs.publish_eval_image
        uses: docker/login-action@dbcb813823bdd20940b903addbd779551569679f # v4
        with:
          registry: ghcr.io
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}
      - name: Build checksum-verified Grok provider image
        if: inputs.publish_eval_image
        env:
          SOURCE_SHA: ${{ needs.authorize_manual.outputs.target_sha }}
        run: |
          set -euo pipefail
          image="ghcr.io/paperclipai/paperclip-daytona-runner:qualification-${SOURCE_SHA}-${GITHUB_RUN_ID}"
          lock_sha="$(sha256sum pnpm-lock.yaml | cut -d ' ' -f 1)"
          docker buildx build --platform linux/amd64 \
            --file docker/daytona-runner/Dockerfile \
            --build-arg "PAPERCLIP_RUNNER_SOURCE_REVISION=$SOURCE_SHA" \
            --build-arg "PAPERCLIP_RUNNER_CONTENT_ID=qualification-$SOURCE_SHA" \
            --build-arg "PAPERCLIP_RUNNER_LOCK_SHA256=$lock_sha" \
            --cache-from type=registry,ref=ghcr.io/paperclipai/paperclip-daytona-runner:e2e-buildcache-amd64 \
            --tag "$image" --metadata-file remote-verification/image.json --push .
          digest="$(jq -r '."containerimage.digest"' remote-verification/image.json)"
          [[ "$digest" =~ ^sha256:[0-9a-f]{64}$ ]]
          immutable="${image%:*}@$digest"
          docker logout ghcr.io
          docker buildx imagetools inspect "$immutable" >/dev/null
          echo "$immutable" > remote-verification/image.txt
          echo "Image: $immutable" >> "$GITHUB_STEP_SUMMARY"
      - name: Verify repository on EC2
        if: inputs.verify_source
        # Leave time for artifact retention before the fleet's one-hour lifetime.
        timeout-minutes: 38
        run: |
          set -uo pipefail
          pnpm install --frozen-lockfile --ignore-scripts
          status=0
          for check in 'pnpm -r typecheck' 'pnpm test:run' 'pnpm check:token-gates' 'pnpm test:e2e:runner:typecheck' 'pnpm test:e2e:runner:unit' 'pnpm build' 'if [ -f scripts/verify-grok-npm-install.mjs ]; then node scripts/verify-grok-npm-install.mjs; fi'; do
            label="$(echo "$check" | tr -cs 'a-zA-Z0-9' '-')"
            echo "Starting $check"
            check_status=0
            timeout --signal=TERM --kill-after=15s 15m bash -c "$check" > "remote-verification/$label.log" 2>&1 || check_status=$?
            printf '%s\t%s\n' "$check_status" "$check" >> remote-verification/check-status.tsv
            if [ "$check_status" -ne 0 ]; then status=1; fi
            echo "Finished $check (exit $check_status)"
          done
          exit "$status"
      - name: Verify clean public npm installation
        if: inputs.verify_public_install && !inputs.verify_source
        timeout-minutes: 35
        run: |
          set -euo pipefail
          test -f scripts/verify-grok-npm-install.mjs || { echo "Selected source does not provide the public-install verifier"; exit 1; }
          pnpm install --frozen-lockfile --ignore-scripts > remote-verification/npm-setup.log 2>&1
          pnpm build > remote-verification/npm-build.log 2>&1
          node scripts/verify-grok-npm-install.mjs > remote-verification/public-npm-install.log 2>&1
      - name: Build canonical eval report viewer
        if: always() && (inputs.verify_source || inputs.build_eval_viewer)
        run: |
          set -euo pipefail
          pnpm install --frozen-lockfile --ignore-scripts --filter '@paperclipai/paperclip-runner...'
          pnpm --filter @paperclipai/paperclip-runner build:issue-thread > remote-verification/viewer-build.log 2>&1
          tar -czf remote-verification/eval-viewer.tar.gz -C packages/paperclip-runner dist-issue-thread
          sha256sum remote-verification/eval-viewer.tar.gz > remote-verification/eval-viewer.sha256
      - name: Retain source, image and verification evidence
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: grok-remote-verification-${{ github.run_id }}
          path: remote-verification/
          retention-days: 7