autofirmer-expanded/extension/background.js

/* AutoFirmer Capture — MV3 service worker.
 *
 * Loop: every tick, scrape the tab you are looking at and POST it to
 * /api/autobuyer/capture. The dashboard's AutoBuyer page renders whatever
 * landed there.
 *
 * Nothing here has a switch. Capture is scoped instead: only the ACTIVE tab of
 * the last-focused window, and only when that tab is one of the firm sites in
 * host_permissions. Sit on any other page and nothing is read or sent.
 *
 * The locate queue is served whenever clicks are waiting. The AutoBuyer is
 * ready at all times; a run happens because someone asked for one.
 */

const DEFAULTS = {
    apiBase: 'http://localhost:3000',
    pollSeconds: 3,
    // Optional CSS selector — matching elements get their on-screen position
    // measured alongside the HTML.
    elementSelector: '',
};

let timer = null;
let pendingLocate = 0;

// ── Config / state ──────────────────────────────────────────────────────────

async function getConfig() {
    const stored = await chrome.storage.local.get(Object.keys(DEFAULTS));
    const cfg = { ...DEFAULTS, ...stored };
    cfg.apiBase = String(cfg.apiBase).replace(/\/+$/, '');
    return cfg;
}

/** Mirror the worker's status into storage so the popup can render it. */
async function setState(patch) {
    const { state: prev } = await chrome.storage.local.get('state');
    await chrome.storage.local.set({ state: { ...(prev || {}), ...patch, updatedAt: Date.now() } });
}

/** `!` when the dashboard is unreachable, blank otherwise. A blank badge is the
 *  working state — there is nothing to switch on, so nothing to advertise. */
function setBadge(connected) {
    chrome.action.setBadgeText({ text: connected ? '' : '!' });
    chrome.action.setBadgeBackgroundColor({ color: '#ef4444' });
}

// ── The injected scraper ────────────────────────────────────────────────────

/** Runs in the page's own world. Must be self-contained — nothing from this
 *  file's scope is available inside it. */
function pageCapture(selector) {
    // getBoundingClientRect() measures from the top-left of the viewport, so
    // rect.left / rect.top ARE the element's position relative to the window.
    // + scrollX/scrollY  -> position within the document
    // + screenX/screenY and the chrome height -> absolute desktop coordinates,
    //   which is what an OS-level clicker needs.
    const chromeHeight = window.outerHeight - window.innerHeight;

    const elements = selector
        ? Array.from(document.querySelectorAll(selector)).slice(0, 200).map((el, index) => {
            const r = el.getBoundingClientRect();
            return {
                index,
                tag: el.tagName.toLowerCase(),
                id: el.id || null,
                text: (el.textContent || '').trim().slice(0, 80),
                // Off-screen or display:none elements still return a rect — all zeroes.
                visible: r.width > 0 && r.height > 0,
                inViewport: r.top < window.innerHeight && r.bottom > 0 && r.left < window.innerWidth && r.right > 0,
                viewport: { x: r.left, y: r.top, width: r.width, height: r.height },
                page: { x: r.left + window.scrollX, y: r.top + window.scrollY },
                // Centre of the element in desktop coordinates (CSS pixels —
                // multiply by devicePixelRatio for physical pixels on HiDPI).
                screen: {
                    x: window.screenX + r.left + r.width / 2,
                    y: window.screenY + chromeHeight + r.top + r.height / 2,
                },
            };
        })
        : [];

    return {
        url: location.href,
        title: document.title,
        html: document.documentElement.outerHTML,
        viewport: {
            scrollX: window.scrollX,
            scrollY: window.scrollY,
            innerWidth: window.innerWidth,
            innerHeight: window.innerHeight,
            outerWidth: window.outerWidth,
            outerHeight: window.outerHeight,
            screenX: window.screenX,
            screenY: window.screenY,
            chromeHeight,
            devicePixelRatio: window.devicePixelRatio,
        },
        elements,
    };
}

// ── Capture pipeline ────────────────────────────────────────────────────────

/** The sites this extension is allowed to touch, taken from the manifest rather
 *  than from a setting.
 *
 *  These are the same hosts host_permissions grants, so they cannot drift out of
 *  sync with what the extension can actually read, and adding a firm — which
 *  means adding its host to the manifest anyway — scopes capture automatically.
 *  localhost is filtered out: that is the dashboard, and capturing it would just
 *  mirror our own output back. */
function firmPatterns() {
    return chrome.runtime.getManifest().host_permissions
        .filter((p) => !/\/\/(localhost|127\.0\.0\.1)/.test(p));
}

/** The firm tab you are actually looking at, or nothing.
 *
 *  Deliberately no fallback to "some other firm tab that happens to be open":
 *  capture should report the page in front of you, not a stale background tab
 *  you left on a broker site hours ago. Chrome does the pattern matching, so a
 *  tab outside host_permissions cannot be selected here at all.
 */
async function pickTab() {
    const [tab] = await chrome.tabs.query({
        active: true,
        lastFocusedWindow: true,
        url: firmPatterns(),
    });
    return tab && /^https?:/.test(tab.url || '') ? tab : undefined;
}

async function captureAndSend(cfg) {
    const tab = await pickTab();
    if (!tab) {
        await setState({ connected: true, error: null, waiting: 'Not on a firm page — nothing captured' });
        return;
    }

    const [injection] = await chrome.scripting.executeScript({
        target: { tabId: tab.id },
        func: pageCapture,
        args: [cfg.elementSelector || ''],
    });
    const result = injection?.result;
    if (!result) throw new Error('Injection returned nothing');

    const res = await fetch(`${cfg.apiBase}/api/autobuyer/capture`, {
        method: 'POST',
        headers: { 'Content-Type': 'application/json' },
        body: JSON.stringify({
            ...result,
            // Stamp the build so the dashboard side can tell which version of this
            // worker is actually live — Chrome keeps running the old one until the
            // extension is reloaded, which is invisible from the server otherwise.
            viewport: { ...result.viewport, extVersion: chrome.runtime.getManifest().version },
            capturedAt: Date.now(),
        }),
    });
    if (!res.ok) throw new Error(`Capture POST failed: HTTP ${res.status}`);

    await setState({
        connected: true,
        error: null,
        waiting: null,
        lastCapture: { url: result.url, title: result.title, bytes: result.html.length, at: Date.now() },
    });
}

/** Which browser is hosting this extension.
 *
 *  The clicker has to raise *this* browser before clicking, and picking by list
 *  order gets it wrong the moment two are installed — it would raise Edge while
 *  the coordinates came from a tab in Chrome, landing every click in the wrong
 *  window. So the answer travels with the measurement.
 */
function detectBrowser() {
    const ua = navigator.userAgent || '';
    if (/\bEdg\//.test(ua)) return 'edge';
    if (/\bOPR\//.test(ua)) return 'opera';
    if (/\bVivaldi\//.test(ua)) return 'vivaldi';
    try {
        if (navigator.brave) return 'brave';   // Brave otherwise reports as Chrome
    } catch { /* not Brave */ }
    return 'chrome';
}

// ── Locate: turn a CSS selector into desktop coordinates ────────────────────

/** Runs in the page. Scrolls the element into view, then reports where it ended up. */
function pageLocate(selector, index) {
    const matches = document.querySelectorAll(selector);
    if (matches.length === 0) return { error: `No element matches ${selector}` };
    const el = matches[index];
    if (!el) return { error: `Selector ${selector} matched ${matches.length} element(s), no index ${index}` };

    // Centre it in the viewport first — an element scrolled off-screen has a rect,
    // but clicking those coordinates would hit whatever is actually displayed there.
    el.scrollIntoView({ block: 'center', inline: 'center', behavior: 'instant' });

    const r = el.getBoundingClientRect();
    if (r.width === 0 || r.height === 0) {
        return { error: `Element ${selector} has zero size (hidden or display:none)` };
    }
    if (r.bottom <= 0 || r.top >= window.innerHeight || r.right <= 0 || r.left >= window.innerWidth) {
        return { error: `Element ${selector} is outside the viewport even after scrolling` };
    }

    const chromeHeight = window.outerHeight - window.innerHeight;
    const cx = r.left + r.width / 2;
    const cy = r.top + r.height / 2;

    // What sits at that point? If it's not our element (or a descendant), something
    // is covering it — a cookie banner, a modal — and the click would hit that instead.
    const atPoint = document.elementFromPoint(cx, cy);
    const covered = atPoint && atPoint !== el && !el.contains(atPoint);

    // Current contents, so the caller can confirm afterwards that what it typed
    // actually landed — a field that silently ignored the keystrokes (masked,
    // read-only, or never focused) is otherwise indistinguishable from success.
    const isField = el.tagName === 'INPUT' || el.tagName === 'TEXTAREA';
    const value = isField ? el.value : (el.isContentEditable ? el.innerText : null);

    return {
        tag: el.tagName.toLowerCase(),
        text: (el.textContent || '').trim().slice(0, 80),
        matchCount: matches.length,
        value,
        editable: (isField || el.isContentEditable) && !el.disabled && !el.readOnly,
        inputType: isField ? (el.type || null) : null,
        covered: !!covered,
        coveredBy: covered ? `${atPoint.tagName.toLowerCase()}${atPoint.id ? '#' + atPoint.id : ''}` : null,
        viewport: { x: r.left, y: r.top, width: r.width, height: r.height },
        // Desktop coordinates of the element's centre, in CSS pixels.
        screen: { x: window.screenX + cx, y: window.screenY + chromeHeight + cy },
        // Lets the Python side work out whether the OS uses a different pixel
        // scale than CSS does (Windows display scaling, some Linux setups).
        screenSize: { width: window.screen.width, height: window.screen.height },
        devicePixelRatio: window.devicePixelRatio,
        chromeHeight,
        url: location.href,
    };
}

/** Is the tab already showing this page? Compares origin and path only —
 *  query strings and hashes shouldn't force a reload. */
function alreadyAt(current, target) {
    try {
        const a = new URL(current);
        const b = new URL(target);
        return a.origin === b.origin && a.pathname.replace(/\/$/, '') === b.pathname.replace(/\/$/, '');
    } catch {
        return false;
    }
}

/** Resolve after the tab reports `complete`. A tab that has only just been
 *  created has no DOM to inject into yet. */
function waitForTabLoad(tabId, timeoutMs = 15000) {
    return new Promise((resolve, reject) => {
        const started = Date.now();
        const check = async () => {
            let tab;
            try {
                tab = await chrome.tabs.get(tabId);
            } catch {
                return reject(new Error('the tab was closed while loading'));
            }
            if (tab.status === 'complete') return resolve(tab);
            if (Date.now() - started > timeoutMs) {
                return reject(new Error(`the page did not finish loading within ${timeoutMs / 1000}s`));
            }
            setTimeout(check, 200);
        };
        check();
    });
}

/** Runs in the page. Scrolls until the list stops growing, then reports what it
 *  ended up with.
 *
 *  The scrolling element is often NOT the window — lists like this usually live
 *  in a div with its own overflow, and scrolling the document does nothing at
 *  all. So walk up from a matched item looking for the ancestor that actually
 *  scrolls, and let the caller name one outright when the guess is wrong.
 */
async function pageScrollToLoad(selector, containerSelector, maxScrolls, settleMs, stableRounds) {
    const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
    const count = () => document.querySelectorAll(selector).length;

    function scroller() {
        if (containerSelector) {
            const named = document.querySelector(containerSelector);
            if (!named) return { error: `No element matches container ${containerSelector}` };
            return { el: named };
        }

        // An ancestor that can actually scroll: overflow allows it, and there is
        // more content than fits.
        let node = document.querySelector(selector)?.parentElement ?? null;
        while (node && node !== document.body) {
            const overflow = getComputedStyle(node).overflowY;
            if (/(auto|scroll)/.test(overflow) && node.scrollHeight > node.clientHeight + 4) {
                return { el: node };
            }
            node = node.parentElement;
        }
        return { el: document.scrollingElement || document.documentElement, isDocument: true };
    }

    const found = scroller();
    if (found.error) return { error: found.error };
    const el = found.el;

    const before = count();
    let last = before;
    let stable = 0;
    let scrolls = 0;

    while (scrolls < maxScrolls) {
        const wasAtBottom = el.scrollTop + el.clientHeight >= el.scrollHeight - 4;
        el.scrollTop = el.scrollHeight;
        scrolls++;
        await sleep(settleMs);

        const now = count();
        // Nothing new AND we were already pinned to the bottom: the list is done
        // growing, not merely slow.
        stable = (now > last) ? 0 : stable + (wasAtBottom ? 1 : 0);
        last = now;
        if (stable >= stableRounds) break;
    }

    return {
        before,
        after: last,
        scrolls,
        exhausted: stable >= stableRounds,
        container: found.isDocument ? 'document' : (el.className || el.tagName || 'element').toString().slice(0, 60),
        url: location.href,
    };
}

async function resolveLocateTab(cfg, request) {
    // Every step carries its firm's pattern; the manifest hosts are the fallback
    // for a bare request (the CLI's locate without --url).
    const pattern = request.urlPattern || firmPatterns();
    if (pattern) {
        const tabs = await chrome.tabs.query({ url: pattern });
        const tab = tabs.find((t) => /^https?:/.test(t.url || ''));
        if (tab) return { tab, opened: false };

        // Nothing matching is open. Open it rather than failing the run — but only
        // to the URL the firm declared, never to something a request supplied that
        // we have no host permission for.
        if (request.openUrl) {
            const created = await chrome.tabs.create({ url: request.openUrl, active: true });
            await waitForTabLoad(created.id);
            return { tab: await chrome.tabs.get(created.id), opened: true };
        }

        throw new Error(`No open tab matches ${pattern}, and no URL is configured to open`);
    }

    const [tab] = await chrome.tabs.query({ active: true, lastFocusedWindow: true });
    if (!tab || !/^https?:/.test(tab.url || '')) throw new Error('No eligible active tab');
    return { tab, opened: false };
}

/** Returns true if a request was handled, false if the queue was empty. */
async function serveLocateRequest(cfg) {
    const claim = await fetch(`${cfg.apiBase}/api/autobuyer/locate/claim`, { method: 'POST' });
    if (!claim.ok) throw new Error(`Claim failed: HTTP ${claim.status}`);
    const { request } = await claim.json();
    if (!request) return false;

    let result = null;
    let error = null;
    try {
        let { tab, opened } = await resolveLocateTab(cfg, request);

        // A navigate step points the tab at a specific page first. Skip it when we
        // are already there — reloading would throw away page state for nothing,
        // and the common case is that the tab is on the right page already.
        if (request.navigateUrl && !alreadyAt(tab.url, request.navigateUrl)) {
            await chrome.tabs.update(tab.id, { url: request.navigateUrl });
            await waitForTabLoad(tab.id);
            tab = await chrome.tabs.get(tab.id);
            opened = true;   // treat as a cold load: the app still has to mount
        }

        // The reported coordinates are only worth anything if that tab is the one
        // actually visible: raise its window and bring the tab to the front.
        // Only un-minimise — passing state:'normal' unconditionally would drop a
        // maximised or fullscreen window out of that state on every single click.
        const win = await chrome.windows.get(tab.windowId);
        const focus = win.state === 'minimized'
            ? { focused: true, state: 'normal' }
            : { focused: true };
        await chrome.windows.update(tab.windowId, focus);
        await chrome.tabs.update(tab.id, { active: true });
        await new Promise((r) => setTimeout(r, 250));   // let the OS finish raising it

        // A scroll request is a different operation on the same channel: no
        // element is measured, the page is just walked to the bottom.
        if (request.options && request.options.op === 'scrollToLoad') {
            const [scrolled] = await chrome.scripting.executeScript({
                target: { tabId: tab.id },
                func: pageScrollToLoad,
                args: [
                    request.selector,
                    request.options.containerSelector || '',
                    Number(request.options.maxScrolls) || 25,
                    Number(request.options.settleMs) || 800,
                    Number(request.options.stableRounds) || 2,
                ],
            });
            const out = scrolled?.result;
            if (!out) throw new Error('Scroll injection returned nothing');
            if (out.error) throw new Error(out.error);
            result = { ...out, tabId: tab.id, windowId: tab.windowId, browser: detectBrowser() };

            await fetch(`${cfg.apiBase}/api/autobuyer/locate/result`, {
                method: 'POST',
                headers: { 'Content-Type': 'application/json' },
                body: JSON.stringify({ id: request.id, result, error: null }),
            });
            return true;
        }

        // `complete` only means the document loaded — a React app still has to
        // mount and paint. Retry briefly rather than declaring the element missing,
        // with a longer budget when we just opened the page from cold.
        const deadline = Date.now() + (opened ? 8000 : 2500);
        let out;
        for (;;) {
            const [injection] = await chrome.scripting.executeScript({
                target: { tabId: tab.id },
                func: pageLocate,
                args: [request.selector, request.index || 0],
            });
            out = injection?.result;
            if (!out) throw new Error('Injection returned nothing');
            if (!out.error) break;
            if (Date.now() >= deadline) throw new Error(out.error);
            await new Promise((r) => setTimeout(r, 350));
        }
        result = {
            ...out,
            tabId: tab.id,
            windowId: tab.windowId,
            openedTab: opened,
            browser: detectBrowser(),
        };
    } catch (err) {
        error = err.message;
    }

    await fetch(`${cfg.apiBase}/api/autobuyer/locate/result`, {
        method: 'POST',
        headers: { 'Content-Type': 'application/json' },
        body: JSON.stringify({ id: request.id, result, error }),
    });
    return true;
}

async function tick() {
    const cfg = await getConfig();

    try {
        const res = await fetch(`${cfg.apiBase}/api/autobuyer/status`, { cache: 'no-store' });
        if (!res.ok) throw new Error(`HTTP ${res.status}`);
        const status = await res.json();
        pendingLocate = Number(status.pendingLocate) || 0;
    } catch (err) {
        setBadge(false);
        await setState({ connected: false, error: `Cannot reach ${cfg.apiBase} — ${err.message}` });
        return;
    }

    setBadge(true);
    await setState({ connected: true });

    // Serve queued clicks before capturing — the Python side is waiting on these,
    // and a capture of a megabyte page shouldn't sit in front of them.
    // Failures inside a request are reported back to the caller; a failure of the
    // queue plumbing itself lands here, and must not take the capture loop down.
    try {
        for (let i = 0; i < pendingLocate; i++) {
            if (!(await serveLocateRequest(cfg))) break;
        }
    } catch (err) {
        await setState({ connected: true, error: `Locate queue: ${err.message}` });
    }

    try {
        await captureAndSend(cfg);
    } catch (err) {
        await setState({ connected: true, error: err.message });
    }
}

// ── Scheduling ──────────────────────────────────────────────────────────────

async function restart() {
    if (timer) clearInterval(timer);
    const cfg = await getConfig();
    // Re-check after the await: restart() is called from several places at startup,
    // and without this a concurrent call leaks an untracked interval — two polling
    // loops in one worker, capturing everything twice.
    if (timer) clearInterval(timer);
    const ms = Math.max(1, Number(cfg.pollSeconds) || DEFAULTS.pollSeconds) * 1000;
    timer = setInterval(tick, ms);
    tick();
}

chrome.runtime.onInstalled.addListener(restart);
chrome.runtime.onStartup.addListener(restart);

// The setInterval above only survives while the worker is alive. Each tick makes
// an extension API call, which resets the idle timer, so in practice the loop
// keeps itself running — this alarm is the safety net that revives it if Chrome
// tears the worker down anyway. 30s is the shortest period Chrome honours.
chrome.alarms.create('keepalive', { periodInMinutes: 0.5 });
chrome.alarms.onAlarm.addListener(() => {
    if (timer) tick(); else restart();
});

chrome.storage.onChanged.addListener((changes, area) => {
    if (area !== 'local') return;
    if (Object.keys(DEFAULTS).some((k) => k in changes)) restart();
});

chrome.runtime.onMessage.addListener((msg, _sender, sendResponse) => {
    if (msg?.type !== 'capture-now') return;
    getConfig()
        .then(captureAndSend)
        .then(() => sendResponse({ ok: true }))
        .catch((err) => sendResponse({ ok: false, error: err.message }));
    return true;   // keep the message channel open for the async response
});

restart();