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

name: Docker Runner check

on:
  workflow_dispatch:
    inputs:
      target_branch:
        description: "Branch or full commit SHA in this repository to build"
        type: string
        required: false
      build_eval_image:
        description: "Build an immutable full-source qualification OCI artifact without publication"
        type: boolean
        default: false
      publish_eval_image:
        description: "Publish an immutable Daytona qualification image on hosted Linux (no provider credentials)"
        type: boolean
        default: false
      release_run_id:
        description: "Consume a successful exact-source Release Smoke pack, daemon and public-package artifact set"
        type: string
        required: false
      release_base_image:
        description: "Immutable Paperclip Daytona base image for the release-artifact consumer"
        type: string
        required: false
      verify_source:
        description: "Run broad source checks on hosted Linux"
        type: boolean
        default: false
      verify_public_install:
        description: "Build and verify clean public npm installation on hosted Linux (no provider credentials)"
        type: boolean
        default: false
      build_eval_viewer:
        description: "Build the canonical eval report viewer on hosted Linux"
        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.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 }}
          RELEASE_RUN_ID: ${{ inputs.release_run_id }}
          PUBLISH_EVAL_IMAGE: ${{ inputs.publish_eval_image }}
          VERIFY_SOURCE: ${{ inputs.verify_source }}
          VERIFY_PUBLIC_INSTALL: ${{ inputs.verify_public_install }}
          BUILD_EVAL_VIEWER: ${{ inputs.build_eval_viewer }}
        run: |
          set -euo pipefail
          test "$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' <<< "$ALLOWED_IDS" >/dev/null
          done
          if [ -n "$RELEASE_RUN_ID" ]; then
            [[ "$RELEASE_RUN_ID" =~ ^[1-9][0-9]*$ ]]
            test "$PUBLISH_EVAL_IMAGE" = true
            test "$VERIFY_SOURCE" != true
            test "$VERIFY_PUBLIC_INSTALL" != true
            test "$BUILD_EVAL_VIEWER" != true
          fi
          if [[ "$TARGET_BRANCH" =~ ^[a-f0-9]{40}$ ]]; then
            target_sha="$(gh api "repos/$REPOSITORY/commits/$TARGET_BRANCH" --jq .sha)"
          else
            target_sha="$(gh api "repos/$REPOSITORY/git/ref/heads/$TARGET_BRANCH" --jq '.object.sha')"
          fi
          [[ "$target_sha" =~ ^[0-9a-f]{40}$ ]]
          echo "target_sha=$target_sha" >> "$GITHUB_OUTPUT"

  manual_image:
    name: Build and verify on Linux
    needs: authorize_manual
    # Keep maintainer authorization below; use a native hosted Linux builder.
    runs-on: ubuntu-latest
    timeout-minutes: 90
    permissions:
      contents: read
      actions: 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: ${{ inputs.release_run_id != '' && github.sha || 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: Apply reviewed qualification lock corrections
        if: inputs.release_run_id == '' && (inputs.build_eval_image || inputs.publish_eval_image)
        env:
          TARGET_SHA: ${{ needs.authorize_manual.outputs.target_sha }}
        run: |
          set -euo pipefail
          if [ "$TARGET_SHA" = "$GITHUB_SHA" ]; then
            node tests/runner-e2e/prepare-report-lock.mjs
          fi

      - name: Resolve source dependencies without lifecycle scripts
        if: inputs.release_run_id == ''
        run: |
          if [[ "${{ inputs.build_eval_image || inputs.publish_eval_image }}" == true ]]; then
            pnpm install --frozen-lockfile --ignore-scripts
          else
            pnpm install --resolution-only --ignore-scripts --no-frozen-lockfile
          fi
          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: ${{ secrets.GITHUB_TOKEN }}
      - name: Prepare hosted image build capacity
        if: (inputs.build_eval_image || inputs.publish_eval_image) && inputs.release_run_id == ''
        run: |
          # These preinstalled SDKs are absent from the image build context.
          sudo rm -rf /usr/share/dotnet /usr/local/lib/android /opt/ghc
          df -h .
          test "$(df --output=avail -BG . | tail -1 | tr -dc '0-9')" -ge 35
      - name: Build checksum-verified qualification image
        if: (inputs.build_eval_image || inputs.publish_eval_image) && inputs.release_run_id == ''
        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)"
          content_id="$(node cli/node_modules/tsx/dist/cli.mjs tests/runner-e2e/daytona-image-content.ts)"
          [[ "$content_id" =~ ^[0-9a-f]{64}$ ]]
          printf '%s\n' "$content_id" > remote-verification/content-id.txt
          output_args=(--output "type=oci,dest=remote-verification/image.oci.tar")
          if [[ "${{ inputs.publish_eval_image }}" == true ]]; then
            output_args=(--push)
          fi
          docker buildx build --platform linux/amd64 \
            --file docker/daytona-runner/Dockerfile \
            --build-arg "PAPERCLIP_RUNNER_SOURCE_REVISION=$SOURCE_SHA" \
            --build-arg "PAPERCLIP_RUNNER_CONTENT_ID=$content_id" \
            --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 "${output_args[@]}" .
          digest="$(jq -r '."containerimage.digest"' remote-verification/image.json)"
          [[ "$digest" =~ ^sha256:[0-9a-f]{64}$ ]]
          immutable="${image%:*}@$digest"
          if [[ "${{ inputs.publish_eval_image }}" == true ]]; then
            docker logout ghcr.io
            docker buildx imagetools inspect "$immutable" >/dev/null
          else
            sha256sum remote-verification/image.oci.tar > remote-verification/image-archive.sha256
          fi
          echo "$immutable" > remote-verification/image.txt
          echo "Image: $immutable; published: ${{ inputs.publish_eval_image }}" >> "$GITHUB_STEP_SUMMARY"
      - name: Consume exact release artifacts into the qualification image
        if: inputs.publish_eval_image && inputs.release_run_id != ''
        env:
          GH_TOKEN: ${{ github.token }}
          SOURCE_SHA: ${{ needs.authorize_manual.outputs.target_sha }}
          RELEASE_RUN_ID: ${{ inputs.release_run_id }}
          RELEASE_BASE_IMAGE: ${{ inputs.release_base_image }}
        run: |
          set -euo pipefail
          test '${{ inputs.verify_source }}' != true
          test '${{ inputs.verify_public_install }}' != true
          test '${{ inputs.build_eval_viewer }}' != true
          mkdir -p remote-verification
          git rev-parse HEAD > remote-verification/tooling-source.txt
          printf '%s\n' "$SOURCE_SHA" > remote-verification/source.txt
          node scripts/prepare-runner-release-image.mjs remote-verification/release-context
          image="ghcr.io/paperclipai/paperclip-daytona-runner:qualification-${SOURCE_SHA}-${GITHUB_RUN_ID}"
          docker buildx build --platform linux/amd64 \
            --build-arg "BASE_IMAGE=$RELEASE_BASE_IMAGE" \
            --build-arg "SOURCE_REVISION=$SOURCE_SHA" \
            --tag "$image" --metadata-file remote-verification/image.json --push \
            remote-verification/release-context
          digest="$(jq -r '."containerimage.digest"' remote-verification/image.json)"
          [[ "$digest" =~ ^sha256:[0-9a-f]{64}$ ]]
          immutable="${image%:*}@$digest"
          docker run --rm --entrypoint cat "$immutable" /opt/paperclip-runner/provider-pack/provider-pack.json \
            > remote-verification/image-provider-pack.json
          cmp remote-verification/release-context/provider-pack.json remote-verification/image-provider-pack.json
          test "$(docker run --rm --entrypoint id "$immutable" -un)" = daytona
          docker run --rm --entrypoint sha256sum "$immutable" /usr/local/bin/paperclip-runnerd \
            > remote-verification/image-daemon.sha256
          expected="$(jq -r .daemonSha256 remote-verification/release-image-inputs.json)"
          test "$(cut -d ' ' -f 1 remote-verification/image-daemon.sha256)" = "$expected"
          printf '%s\n' "$immutable" > remote-verification/image.txt
          rm -rf remote-verification/release-context
          docker logout ghcr.io
          echo "Image: $immutable" >> "$GITHUB_STEP_SUMMARY"
      - name: Verify repository on hosted Linux
        if: inputs.verify_source
        # Leave time for artifact retention within the job deadline.
        timeout-minutes: 80
        run: |
          set -uo pipefail
          pnpm install --frozen-lockfile --ignore-scripts
          git fetch --no-tags --depth=1 origin 2a8a99e4a5f69aa803b3f10b982f583e75a87042 16ef1a5744b23bb043a8b286e6288921da1de98f
          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 $check"
            check_status=0
            limit=15m
            if [ "$check" = 'pnpm test:run' ]; then limit=60m; fi
            timeout --signal=TERM --kill-after=15s "$limit" bash -c "$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 --resolution-only --ignore-scripts --no-frozen-lockfile > remote-verification/npm-setup.log 2>&1
          pnpm install --frozen-lockfile --ignore-scripts >> remote-verification/npm-setup.log 2>&1
          pnpm build > remote-verification/npm-build.log 2>&1
          PAPERCLIP_RUNNER_QUALIFICATION_PACKAGES_DIR="$PWD/remote-verification/qualification-packages" \
            node scripts/verify-grok-npm-install.mjs > remote-verification/public-npm-install.log 2>&1
      - name: Build canonical eval report viewer
        if: always() && (inputs.verify_source || inputs.build_eval_viewer)
        run: |
          set -euo pipefail
          pnpm install --frozen-lockfile --ignore-scripts --filter '@paperclipai/paperclip-runner...'
          pnpm --filter @paperclipai/paperclip-runner build:issue-thread > remote-verification/viewer-build.log 2>&1
          tar -czf remote-verification/eval-viewer.tar.gz -C packages/paperclip-runner dist-issue-thread
          sha256sum remote-verification/eval-viewer.tar.gz > remote-verification/eval-viewer.sha256
      - name: Retain source, image and verification evidence
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: grok-remote-verification-${{ github.run_id }}
          path: |
            remote-verification/
            !remote-verification/release-context/**
          compression-level: 0
          retention-days: ${{ inputs.build_eval_image && 1 || 7 }}