Files
PaperClipAI/.github/workflows/docker-runner-check.yml
T
2026-10-09 01:59:03 -05:00

268 lines
13 KiB
YAML

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
candidate_provider:
description: "Include a pending native harness in the qualification image"
type: choice
options:
- none
- hermes
default: none
verify_source:
description: "Run broad source checks on EC2"
type: boolean
default: 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.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: ${{ inputs.target_branch || github.ref_name }}
run: |
set -euo pipefail
test "$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')"
[[ "$target_sha" =~ ^[0-9a-f]{40}$ ]]
echo "target_sha=$target_sha" >> "$GITHUB_OUTPUT"
manual_image:
name: Build and verify on EC2
needs: authorize_manual
runs-on: runs-on/fleet=paperclip-public-pr-x64/env=public-ci
timeout-minutes: ${{ inputs.publish_eval_image && !inputs.verify_source && !inputs.verify_public_install && !inputs.build_eval_viewer && 45 || 90 }}
permissions:
contents: read
packages: write
id-token: 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 }}
- if: inputs.publish_eval_image
uses: sigstore/cosign-installer@398d4b0eeef1380460a10c8013a76f728fb906ac # v3
- name: Build checksum-verified qualification image
if: inputs.publish_eval_image
env:
SOURCE_SHA: ${{ needs.authorize_manual.outputs.target_sha }}
CANDIDATE_PROVIDER: ${{ inputs.candidate_provider || 'none' }}
run: |
set -euo pipefail
case "$CANDIDATE_PROVIDER" in
none) candidate_providers='' ;;
hermes) candidate_providers=hermes ;;
*) echo 'Unsupported qualification provider' >&2; exit 1 ;;
esac
image="ghcr.io/paperclipai/paperclip-daytona-runner:qualification-${SOURCE_SHA}-${GITHUB_RUN_ID}"
lock_sha="$(sha256sum pnpm-lock.yaml | cut -d ' ' -f 1)"
content_id="$(printf '%s\n' "$SOURCE_SHA" "$lock_sha" "$candidate_providers" | sha256sum | cut -d ' ' -f 1)"
timeout --signal=TERM --kill-after=15s 40m docker buildx build --platform linux/amd64 \
--file docker/daytona-runner/Dockerfile \
--build-arg "PAPERCLIP_RUNNER_SOURCE_REVISION=$SOURCE_SHA" \
--build-arg "PAPERCLIP_RUNNER_CONTENT_ID=$content_id" \
--build-arg "PAPERCLIP_RUNNER_LOCK_SHA256=$lock_sha" \
--build-arg "PAPERCLIP_RUNNER_CANDIDATE_PROVIDERS=$candidate_providers" \
--cache-from type=registry,ref=ghcr.io/paperclipai/paperclip-daytona-runner:e2e-buildcache-amd64 \
--tag "$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"
cosign sign --yes "$immutable"
cosign verify \
--certificate-identity "https://github.com/$GITHUB_WORKFLOW_REF" \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
"$immutable" > remote-verification/image-signature.json
docker logout ghcr.io
docker buildx imagetools inspect "$immutable" >/dev/null
docker buildx imagetools inspect "$immutable" --format '{{json .Image}}' \
| jq --arg source "$SOURCE_SHA" --arg content "$content_id" -e '
{architecture, os, user: .config.User,
contentId: .config.Labels["io.paperclip.runner.content-id"],
sourceRevision: .config.Labels["org.opencontainers.image.revision"]}
| select(.architecture == "amd64" and .os == "linux" and .user == "daytona"
and .contentId == $content and .sourceRevision == $source)
' > remote-verification/image-identity.json
# Read only public package metadata as the image's unprivileged user.
# Host namespaces and actual Daytona execution need separate proof.
docker run --rm --network none \
--entrypoint /opt/paperclip-runner/provider-pack/node_modules/node/bin/node \
"$immutable" -e '
const fs = require("node:fs");
const assert = require("node:assert/strict");
const pack = JSON.parse(fs.readFileSync("/opt/paperclip-runner/provider-pack/provider-pack.json", "utf8"));
const candidate = process.argv[1];
assert.equal(pack.schema, "paperclip-runner/remote-provider-pack/v1");
assert.equal(pack.payload.runnerSourceRevision, process.argv[2]);
assert.deepEqual(pack.payload.target, {platform: "linux", architecture: "x64"});
assert.match(pack.digest, /^sha256:[a-f0-9]{64}$/);
if (candidate === "hermes") {
const hermes = pack.payload?.candidateProviders?.hermes;
assert.equal(hermes?.version, "v2026.9.24");
assert.equal(hermes?.qualification, "pending");
assert.match(hermes.closureDigest, /^sha256:[a-f0-9]{64}$/);
assert.equal(hermes.path, "provider-assets/hermes/linux-x64");
}
console.log(JSON.stringify({providerPackDigest: pack.digest, candidateProvider: candidate,
candidate: candidate === "hermes" ? pack.payload.candidateProviders.hermes : null}));
' "$CANDIDATE_PROVIDER" "$SOURCE_SHA" > remote-verification/provider-pack-identity.json
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