Files
Stirling-PDF/.taskfiles/e2e.yml
T
Anthony StirlingandGitHub 3ecd95b779 Add MCP server with OAuth/API-key auth (#6570)
Adds an optional MCP server (proprietary module) that exposes Stirling's
PDF operations and AI capabilities to MCP clients. Off by default, zero
footprint when disabled.

### What
- New `/mcp` endpoint: streamable-HTTP + JSON-RPC 2.0; 8 tools
(describe_operation, pages/convert/misc/security category tools, AI,
upload, download).
- Runs real operations over an internal loopback; results returned
inline as base64 (small) or by fileId (large).

### Auth (two modes)
- OAuth2 resource server: RFC 9728 protected-resource metadata, RFC 8707
audience binding, JWKS, `mcp.tools.read/write` scopes; binds each token
to a provisioned Stirling account.
- API-key mode: reuses Stirling per-user `X-API-KEY` (no IdP needed).

### Security
- Per-user file ownership in FileStorage: async/queued writes scoped to
the submitting user; legacy/owner-less files stay readable.
- Admin allow/block list controls which operations are exposed.
- Python engine gated behind a shared secret (`X-Engine-Auth`).
- MCP filter chain is isolated and cannot weaken the main app's
security.
- Hardened: no upstream error-body leakage, log injection sanitized,
fileId path/sidecar enumeration blocked.

### Config / footprint
- Disabled by default (`mcp.enabled=false`); all beans
`@ConditionalOnProperty`.
---

## Checklist

### General

- [ ] I have read the [Contribution
Guidelines](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/CONTRIBUTING.md)
- [ ] I have read the [Stirling-PDF Developer
Guide](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/DeveloperGuide.md)
(if applicable)
- [ ] I have read the [How to add new languages to
Stirling-PDF](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/devGuide/HowToAddNewLanguage.md)
(if applicable)
- [ ] I have performed a self-review of my own code
- [ ] My changes generate no new warnings

### Documentation

- [ ] I have updated relevant docs on [Stirling-PDF's doc
repo](https://github.com/Stirling-Tools/Stirling-Tools.github.io/blob/main/docs/)
(if functionality has heavily changed)
- [ ] I have read the section [Add New Translation
Tags](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/devGuide/HowToAddNewLanguage.md#add-new-translation-tags)
(for new translation tags only)

### Translations (if applicable)

- [ ] I ran
[`scripts/counter_translation.py`](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/docs/counter_translation.md)

### UI Changes (if applicable)

- [ ] Screenshots or videos demonstrating the UI changes are attached
(e.g., as comments or direct attachments in the PR)

### Testing (if applicable)

- [ ] I have run `task check` to verify linters, typechecks, and tests
pass
- [ ] I have tested my changes locally. Refer to the [Testing
Guide](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/DeveloperGuide.md#7-testing)
for more details.
2026-06-10 09:46:25 +00:00

267 lines
10 KiB
YAML

version: '3'
tasks:
install:
desc: "Install Playwright browsers"
dir: frontend/editor
deps: [ ':frontend:install' ]
cmds:
- npx playwright install {{.CLI_ARGS}} --with-deps
stubbed:
desc: "Run stubbed E2E tests"
dir: frontend/editor
deps: [ ':frontend:prepare' ]
cmds:
- npx playwright test --project=stubbed {{.CLI_ARGS}}
live:
desc: "Run live E2E tests"
summary: |
Auto-spawns a Spring Boot backend with isolated state under
.test-state/playwright/ (purged on every run) so the suite doesn't
touch your local DB, settings.yml, backups or customFiles. The
backend is stopped automatically when the runner exits.
Pass extra Playwright flags via -- :
task e2e:live -- --headed --grep merge
deps:
- live:backend
- live:runner
live:backend:
internal: true
ignore_error: true
vars:
BASE_DIR: '{{.ROOT_DIR}}/.test-state/playwright'
# COVERAGE=1 in the calling environment attaches the JaCoCo agent to
# the bootRun JVM and writes to BASE_DIR/jacoco.exec on shutdown.
# Off by default to keep local dev runs uninstrumented; CI flips it.
env:
STIRLING_BASE_PATH: '{{.BASE_DIR}}'
# Suppress the analytics opt-in modal that fires on first admin login.
# It renders a Mantine overlay that intercepts pointer events and blocks
# FirstLoginSlide from rendering, so the bootstrap spec times out waiting
# for the password-change prompt.
SYSTEM_ENABLEANALYTICS: "false"
# NOTE: SECURITY_INITIALLOGIN_USERNAME/PASSWORD are intentionally NOT set.
# The live-setup project's bootstrap spec performs the real first-login
# flow against the backend's default admin/stirling user, exercising the
# forced-password-change UI and leaving the DB at admin/adminadmin for
# the rest of the live suite. This is both real coverage of the first-
# login flow and a stronger seed than env-var-driven user creation.
cmds:
# Wrapped in `bash -c` because Task's embedded shell (mvdan/sh) doesn't
# support `&` + `$!` + `wait` reliably (gradle was never actually
# backgrounded, so `$!` recorded a stale PID and the runner saw an
# immediate "backend died"). Real bash backgrounds gradle properly and
# gives us a stable PID to clean up.
- |
bash -c '
set -e
rm -rf "{{.BASE_DIR}}"
mkdir -p "{{.BASE_DIR}}"
GRADLE_ARGS=":stirling-pdf:bootRun"
if [ -n "${COVERAGE:-}" ]; then
# copyJacocoAgent is wired as a dependency of bootRun when
# -PjacocoAgent=true, so we do not need to invoke it separately.
GRADLE_ARGS="$GRADLE_ARGS -PjacocoAgent=true -PjacocoExec={{.BASE_DIR}}/jacoco.exec"
echo "JaCoCo coverage enabled, writing to {{.BASE_DIR}}/jacoco.exec"
fi
# Background gradle and record its PID so the runner can clean up
# the exact process tree (wrapper + forked Spring Boot JVM) without
# resorting to fuzzy `pkill -f` patterns. `wait` keeps this script
# alive for the lifetime of gradle so Task'"'"'s parallel deps stay
# synchronised.
bash gradlew $GRADLE_ARGS > "{{.BASE_DIR}}/backend.log" 2>&1 &
GRADLE_PID=$!
echo $GRADLE_PID > "{{.BASE_DIR}}/backend.pid"
wait $GRADLE_PID
'
live:runner:
internal: true
deps: [ ':frontend:prepare' ]
dir: frontend/editor
vars:
BASE_DIR: '{{.ROOT_DIR}}/.test-state/playwright'
cmds:
# Wrapped in `bash -c` because Task's embedded shell (mvdan/sh) only
# supports `EXIT`/`ERR` in `trap`, not `INT`/`TERM`. CLI_ARGS are
# passed positionally so quoted args like --grep "foo bar" survive.
- |
bash -c '
set +e
# bootRun forks a separate Spring Boot JVM as a child of the gradle
# wrapper; killing only the wrapper leaves that JVM orphaned holding
# :8080. Walk the recorded PID'"'"'s process tree via pgrep so we kill
# every descendant, then lsof as a last-resort backstop.
kill_tree() {
local pid=$1
for child in $(pgrep -P "$pid" 2>/dev/null); do
kill_tree "$child"
done
kill "$pid" 2>/dev/null || true
}
cleanup() {
if [ -f "{{.BASE_DIR}}/backend.pid" ]; then
echo "Stopping backend..."
kill_tree "$(cat "{{.BASE_DIR}}/backend.pid")"
sleep 1
lsof -ti:8080 2>/dev/null | xargs kill 2>/dev/null || true
echo "Backend stopped"
fi
}
# Trap so cleanup runs on every exit path: tests pass, tests fail,
# backend never starts, user Ctrl-Cs. Without this the gradle JVM
# is left orphaned holding :8080 whenever the wait loop times out.
trap cleanup EXIT INT TERM
echo "Waiting for backend on :8080 (up to 10min; first compile can be slow)..."
WAITED=0
while [ $WAITED -lt 600 ]; do
if curl -sf http://localhost:8080/api/v1/info/status 2>/dev/null | grep -q UP; then
echo "Backend ready, starting Playwright"
break
fi
# Bail early if the backend already died: no point waiting 10min.
if [ -f "{{.BASE_DIR}}/backend.pid" ]; then
BACKEND_PID=$(cat "{{.BASE_DIR}}/backend.pid")
if ! kill -0 "$BACKEND_PID" 2>/dev/null; then
echo "Backend process exited before becoming ready; last 100 log lines:"
tail -100 "{{.BASE_DIR}}/backend.log" 2>/dev/null || true
exit 1
fi
fi
sleep 2
WAITED=$((WAITED + 2))
done
if [ $WAITED -ge 600 ]; then
echo "Backend did not become ready within 10min, aborting"
exit 1
fi
npx playwright test --project=live "$@"
' bash {{.CLI_ARGS}}
enterprise:
desc: "Run enterprise E2E tests"
summary: |
Requires an already-running keycloak compose stack on :8080. Bring one
up first with:
task e2e:oauth:up
task e2e:saml:up
dir: frontend/editor
deps: [ ':frontend:prepare' ]
cmds:
- npx playwright test --project=enterprise {{.CLI_ARGS}}
cross-browser:
desc: "Run stubbed E2E tests on Chromium, Firefox, and WebKit"
dir: frontend/editor
deps: [ ':frontend:prepare' ]
cmds:
- npx playwright test --project=stubbed --project=stubbed-firefox --project=stubbed-webkit {{.CLI_ARGS}}
check:
desc: "Run all E2E tests not requiring Docker"
cmds:
- task: stubbed
- task: live
check:all:
desc: "Run all E2E tests"
summary: |
Includes the enterprise project, which requires a keycloak compose
stack on :8080. Bring one up first with:
task e2e:oauth:up
task e2e:saml:up
cmds:
- task: stubbed
- task: live
- task: enterprise
oauth:up:
desc: "Start the OAuth keycloak test environment"
summary: |
Set LICENSE_KEY=<KEY> to skip the interactive license prompt:
task e2e:oauth:up LICENSE_KEY=abc123
Pass extra flags via -- :
task e2e:oauth:up -- --auto --nobuild
ignore_error: true
cmds:
- bash testing/compose/start-oauth-test.sh {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
oauth:down:
desc: "Stop the OAuth keycloak test environment"
cmds:
- docker compose -f testing/compose/docker-compose-keycloak-oauth.yml down -v
saml:up:
desc: "Start the SAML keycloak test environment"
summary: |
Set LICENSE_KEY=<KEY> to skip the interactive license prompt:
task e2e:saml:up LICENSE_KEY=abc123
Pass extra flags via -- :
task e2e:saml:up -- --auto --with-storage --nobuild --language sv-SE
ignore_error: true
cmds:
- bash testing/compose/start-saml-test.sh {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
saml:down:
desc: "Stop the SAML keycloak test environment"
cmds:
- docker compose -f testing/compose/docker-compose-keycloak-saml.yml down -v
mcp:up:
desc: "Start the MCP keycloak test environment (Stirling as OAuth resource server)"
summary: |
Brings up Keycloak (OAuth authorization server) + Stirling configured as an
MCP resource server, then you can exercise /mcp with real Keycloak tokens.
Set LICENSE_KEY=<KEY> to skip the interactive license prompt:
task e2e:mcp:up LICENSE_KEY=abc123
Pass extra flags via -- :
task e2e:mcp:up -- --validate --nobuild
ignore_error: true
cmds:
- bash testing/compose/start-mcp-test.sh {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
mcp:manual:
desc: "Start the MCP keycloak test env in manual mode (prints URLs + a live token for your client)"
summary: |
Brings the stack up and prints copy-paste URLs/commands plus a freshly minted
access token so you can drive your own MCP client (Inspector, curl, ...).
task e2e:mcp:manual LICENSE_KEY=<your-license-key>
Add --nobuild if the images are already built:
task e2e:mcp:manual LICENSE_KEY=<your-license-key> -- --nobuild
ignore_error: true
cmds:
- bash testing/compose/start-mcp-test.sh --manual {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
mcp:apikey:
desc: "Start the MCP test env in API-KEY manual mode (no OAuth/IdP): mints a key + prints client settings"
summary: |
Brings Stirling up in apikey auth mode and prints copy-paste client settings with a freshly
minted X-API-KEY - ideal for clients whose OAuth layer can't reach localhost.
task e2e:mcp:apikey LICENSE_KEY=<your-license-key>
Add --nobuild if images are already built:
task e2e:mcp:apikey LICENSE_KEY=<your-license-key> -- --nobuild
ignore_error: true
cmds:
- bash testing/compose/start-mcp-test.sh --apikey {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
mcp:validate:
desc: "Validate the running MCP keycloak test environment end-to-end (oauth mode + real MCP SDK client)"
cmds:
- bash testing/compose/validate-mcp-test.sh
mcp:validate-apikey:
desc: "Validate the MCP server in API-KEY auth mode (mints a key + real MCP SDK client), then restore oauth"
cmds:
- bash testing/compose/validate-mcp-apikey.sh
mcp:down:
desc: "Stop the MCP keycloak test environment"
cmds:
- docker compose -f testing/compose/docker-compose-keycloak-mcp.yml down -v