refactor(core): transition to WPE

This commit is contained in:
2026-07-06 18:10:12 +02:00
parent 2415502d4b
commit e3b4692802
4 changed files with 92 additions and 144 deletions
+7 -6
View File
@@ -1,12 +1,16 @@
FROM node:18-bullseye-slim
FROM node:18-bookworm-slim
# Install dependencies for Chromium and X11
# Install dependencies for WPE WebKit (Cog), X11, and DBus
RUN apt-get update && apt-get install -y \
chromium \
cog \
wpewebkit \
x11-xserver-utils \
wmctrl \
xdotool \
scrot \
jq \
curl \
dbus-x11 \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
@@ -18,7 +22,4 @@ COPY server.js ./
COPY start.sh /start.sh
RUN chmod +x /start.sh
# Provide a default user data directory
RUN mkdir -p /data/chromium && chown -R node:node /data
CMD ["/start.sh"]
+1 -2
View File
@@ -1,14 +1,13 @@
{
"name": "kiosk-api",
"version": "1.0.0",
"description": "Multi-monitor kiosk API based on balena browser block",
"description": "Multi-monitor kiosk API using WPE WebKit",
"main": "server.js",
"scripts": {
"start": "node server.js"
},
"dependencies": {
"express": "^4.18.2",
"puppeteer-core": "^20.0.0",
"body-parser": "^1.20.2"
}
}
+75 -130
View File
@@ -1,9 +1,7 @@
const express = require('express');
const bodyParser = require('body-parser');
const puppeteer = require('puppeteer-core');
const { spawn, exec, execSync } = require('child_process');
const { spawn, execSync, exec } = require('child_process');
const fs = require('fs');
const path = require('path');
const app = express();
app.use(bodyParser.urlencoded({ extended: true }));
@@ -14,14 +12,11 @@ const MONITOR = process.env.MONITOR || 'HDMI-1';
// State variables
let browserProcess = null;
let browser = null;
let page = null;
let autorefreshInterval = null;
let currentUrl = process.env.LAUNCH_URL || 'about:blank';
let isKiosk = process.env.KIOSK !== '0'; // default 1
let isGpu = process.env.GPU !== '0'; // default 1
let chromiumFlags = [];
let currentUrl = process.env.LAUNCH_URL || 'https://duckduckgo.com';
let isKiosk = 1; // Cog is always full-screen/kiosk implicitly
let isGpu = 1; // WPE is inherently hardware-accelerated
function getMonitorGeometry(monitorName) {
try {
@@ -29,7 +24,6 @@ function getMonitorGeometry(monitorName) {
const lines = stdout.split('\n');
for (const line of lines) {
if (line.startsWith(monitorName + ' ') && line.includes(' connected')) {
// Match geometry string like 1920x1080+1920+0
const match = line.match(/(\d+)x(\d+)\+(\d+)\+(\d+)/);
if (match) {
return {
@@ -50,7 +44,6 @@ function getMonitorGeometry(monitorName) {
function ensureMonitorConnected() {
return new Promise(resolve => {
const check = () => {
// Run xrandr --auto to ensure any newly connected displays are configured
try {
execSync('xrandr --auto');
} catch(e) {}
@@ -68,91 +61,70 @@ function ensureMonitorConnected() {
});
}
function buildFlags(geometry) {
console.log(`Monitor ${MONITOR} resolved to position ${geometry.x},${geometry.y} with size ${geometry.width}x${geometry.height}`);
const flags = [
'--no-sandbox',
'--disable-dev-shm-usage',
'--remote-debugging-port=9222',
`--user-data-dir=/data/chromium-${MONITOR}`,
`--window-position=${geometry.x},${geometry.y}`,
`--window-size=${geometry.width},${geometry.height}`,
'--autoplay-policy=no-user-gesture-required'
];
if (isKiosk) {
flags.push('--kiosk');
// To ensure full screen stays on the target monitor
flags.push(`--app=${currentUrl}`);
function positionCogWindow(geometry) {
// Wait for the window to appear, then use wmctrl to move it
console.log('Attempting to position Cog window...');
let attempts = 10;
const tryMove = () => {
try {
// Find the window ID of the Cog browser.
// Cog's window title or class is usually 'cog'
const stdout = execSync('wmctrl -l').toString();
if (stdout.toLowerCase().includes('cog')) {
// Find WID
const line = stdout.split('\n').find(l => l.toLowerCase().includes('cog'));
if (line) {
const wid = line.split(' ')[0];
console.log(`Found Cog window: ${wid}. Positioning to ${geometry.x},${geometry.y}...`);
// Remove maximization first
try { execSync(`wmctrl -i -r ${wid} -b remove,maximized_vert,maximized_horz`); } catch (e) {}
// Move and resize: gravity,x,y,w,h
execSync(`wmctrl -i -r ${wid} -e 0,${geometry.x},${geometry.y},${geometry.width},${geometry.height}`);
console.log('Window positioned successfully.');
return; // Done
}
}
} catch (e) {
console.log('wmctrl check failed', e.message);
}
if (!isGpu) {
flags.push('--disable-gpu');
flags.push('--disable-software-rasterizer');
attempts--;
if (attempts > 0) {
setTimeout(tryMove, 1000);
} else {
flags.push('--enable-gpu');
console.error('Failed to find and position Cog window.');
}
};
setTimeout(tryMove, 2000);
}
chromiumFlags = flags;
return flags;
}
async function startChromium() {
async function startBrowser() {
console.log(`Checking connection status for ${MONITOR}...`);
const geometry = await ensureMonitorConnected();
if (browserProcess) {
console.log('Stopping existing Chromium...');
console.log('Stopping existing Cog...');
browserProcess.kill('SIGTERM');
await new Promise(resolve => setTimeout(resolve, 2000));
}
const flags = buildFlags(geometry);
console.log(`Starting Cog to navigate to ${currentUrl}`);
console.log('Starting Chromium with flags:', flags.join(' '));
// If kiosk is enabled, we pass the URL via --app to enforce window boundaries
// Otherwise we just pass the URL as a positional argument.
const args = [...flags];
if (!isKiosk) {
args.push(currentUrl);
}
browserProcess = spawn('/usr/bin/chromium', args, {
stdio: 'inherit'
// Launch Cog. We use the X11 platform if available.
const args = [currentUrl];
const env = Object.assign({}, process.env, {
COG_PLATFORM_NAME: 'x11',
// Ensures background opacity might work and standard behaviors
});
// Connect Puppeteer with retries
let retries = 15;
const tryConnect = async () => {
try {
console.log('Connecting puppeteer...');
browser = await puppeteer.connect({
browserURL: 'http://127.0.0.1:9222',
defaultViewport: null
browserProcess = spawn('cog', args, {
stdio: 'inherit',
env: env
});
const pages = await browser.pages();
page = pages.length > 0 ? pages[0] : await browser.newPage();
if (!isKiosk) {
console.log(`Navigating to ${currentUrl}`);
await page.goto(currentUrl);
}
console.log('Puppeteer connected successfully!');
} catch (e) {
if (retries > 0) {
console.log(`Puppeteer connect failed, retrying in 2s... (${retries} attempts left)`);
retries--;
setTimeout(tryConnect, 2000);
} else {
console.error('Puppeteer connect error (fatal):', e);
}
}
};
setTimeout(tryConnect, 2000);
// Call position function which will wait for window to exist and move it
positionCogWindow(geometry);
}
// API Endpoints
@@ -161,11 +133,11 @@ app.get('/ping', (req, res) => {
res.status(200).send('OK');
});
app.post('/refresh', async (req, res) => {
if (page) {
await page.reload();
app.post('/refresh', (req, res) => {
try {
execSync('cogctl reload');
res.status(200).send('Refreshed');
} else {
} catch (e) {
res.status(503).send('Browser not ready');
}
});
@@ -179,7 +151,7 @@ app.post('/autorefresh/:interval', (req, res) => {
if (interval > 0 && interval <= 60) {
autorefreshInterval = setInterval(() => {
if (page) page.reload().catch(console.error);
try { execSync('cogctl reload'); } catch(e) {}
}, interval * 1000);
res.status(200).send(`Autorefresh enabled: ${interval}s`);
} else {
@@ -195,70 +167,44 @@ app.get('/url', (req, res) => {
res.status(200).send(currentUrl);
});
app.post('/url', async (req, res) => {
let needRestart = false;
app.post('/url', (req, res) => {
if (req.body.url) currentUrl = req.body.url;
if (req.body.gpu !== undefined) {
const newGpu = req.body.gpu === '1';
if (newGpu !== isGpu) {
isGpu = newGpu;
needRestart = true;
}
}
if (req.body.kiosk !== undefined) {
const newKiosk = req.body.kiosk === '1';
if (newKiosk !== isKiosk) {
isKiosk = newKiosk;
needRestart = true;
}
}
// If we're in kiosk mode, --app flag is used so changing URL requires restart to stay full screen
// If we're not in kiosk mode, we can just navigate the page
if (needRestart || isKiosk) {
await startChromium();
} else if (page) {
await page.goto(currentUrl);
}
// Cogctl can directly tell the browser to open a URL without restarting it
try {
console.log(`Sending cogctl open ${currentUrl}`);
execSync(`cogctl open "${currentUrl}"`);
res.status(200).send(`URL set to ${currentUrl}`);
} catch (e) {
console.error('Failed to set URL via cogctl. Attempting restart...', e.message);
startBrowser().then(() => {
res.status(200).send(`URL set to ${currentUrl} (restarted)`);
});
}
});
app.get('/gpu', (req, res) => {
res.status(200).send(isGpu ? '1' : '0');
res.status(200).send('1'); // Always 1 for WPE
});
app.put('/gpu/:value', async (req, res) => {
const val = req.params.value === '1';
if (val !== isGpu) {
isGpu = val;
await startChromium();
}
res.status(200).send(isGpu ? '1' : '0');
app.put('/gpu/:value', (req, res) => {
res.status(200).send('1'); // Ignored for WPE
});
app.get('/kiosk', (req, res) => {
res.status(200).send(isKiosk ? '1' : '0');
res.status(200).send('1'); // Always 1 for WPE
});
app.put('/kiosk/:value', async (req, res) => {
const val = req.params.value === '1';
if (val !== isKiosk) {
isKiosk = val;
await startChromium();
}
res.status(200).send(isKiosk ? '1' : '0');
app.put('/kiosk/:value', (req, res) => {
res.status(200).send('1'); // Ignored for WPE
});
app.get('/flags', (req, res) => {
res.status(200).json(chromiumFlags);
res.status(200).json(['WPE WebKit (Cog)']);
});
app.get('/version', (req, res) => {
exec('/usr/bin/chromium --version', (err, stdout) => {
exec('cog --version', (err, stdout) => {
if (err) res.status(500).send('Unknown');
else res.status(200).send(stdout.trim());
});
@@ -266,7 +212,6 @@ app.get('/version', (req, res) => {
app.get('/screenshot', (req, res) => {
const file = `/tmp/screenshot-${Date.now()}.png`;
// We use scrot for X11 screenshots.
exec(`scrot ${file}`, (err) => {
if (err) return res.status(500).send('Screenshot failed');
res.sendFile(file);
@@ -274,7 +219,7 @@ app.get('/screenshot', (req, res) => {
});
// Start initially
startChromium();
startBrowser();
app.listen(PORT, () => {
console.log(`Kiosk API running on port ${PORT} for monitor ${MONITOR}`);
+5 -2
View File
@@ -6,10 +6,13 @@ while [ ! -S /tmp/.X11-unix/X0 ]; do
sleep 1
done
echo "X11 display found. Starting API server..."
echo "X11 display found. Initializing D-Bus and starting API server..."
# Disable screen blanking for this display as well just in case
# Disable screen blanking
xset s off -dpms || true
# Initialize a local D-Bus session. This is required for cogctl to communicate with cog.
export DBUS_SESSION_BUS_ADDRESS=$(dbus-daemon --session --print-address --fork)
# Start the Node.js API server
exec node server.js