@apmantza/greedysearch-pi 1.8.7 → 1.8.8

package/CHANGELOG.md

@@ -1,9 +1,28 @@
 # Changelog
 
-## [Unreleased]
+## [1.8.8] — 2026-05-09
+
+### Added
+
+- **`/set-greedy-locale` Pi command** (`index.ts`) — Set default locale for search results (e.g., `/set-greedy-locale de`, `/set-greedy-locale --clear`, `/set-greedy-locale --show`). Saves to `~/.config/greedysearch/config.json`.
+- **Browser lifecycle defense patterns** (`src/search/browser-lifecycle.mjs`, new) — Centralized lifecycle management adopted from open-websearch's robust cross-process browser patterns:
+  - **Structured JSON metadata** (`greedysearch-chrome-metadata.json`) replaces three scattered text files (PID, mode, activity) with a single file tracking `browserPid`, `debugPort`, `tempDir`, `clientPids[]`, `sessionMode`, `lastActivity`, `launchedAt`. Backward-compatible — legacy files still written.
+  - **Process command-line verification** — `verifyBrowserProcess()` checks not just PID alive but that the process command line contains the profile dir and debug port. Prevents PID collision false-positives where a different process reuses the same PID.
+  - **Cross-process launch lock** — `acquireLaunchLock()` uses exclusive-create (`wx` flag) to prevent concurrent `ensureChrome()` calls from racing to launch Chrome. Stale lock recovery after 15s.
+  - **Stale session cleanup** — `cleanupStaleSessions()` runs once per process on first `ensureChrome()`. Scans metadata for dead PIDs, verifies survivors via command line, force-kills orphans, reclaims ghost processes on port 9222.
+  - **Client PID tracking** — `registerClient`/`unregisterClient` track which processes share the Chrome instance.
+- **Mode-specific idle timeouts** (`src/search/chrome.mjs`) — Headless Chrome keeps the aggressive 5-minute idle timeout (`GREEDY_SEARCH_IDLE_TIMEOUT_MINUTES`) since it's cheap to restart. Visible Chrome (explicitly launched for captcha/cookie setup) gets a 60-minute grace period (`GREEDY_SEARCH_VISIBLE_IDLE_TIMEOUT_MINUTES`) to avoid wasting the user's captcha investment. Set either to 0 to disable for that mode.
+
+### Added
+
+- **System command path resolution** (`src/utils/system-cmds.mjs`, new) — `resolveSystemCmd()` resolves `powershell`, `netstat`, `taskkill`, `ps`, `lsof`, `ss`, `grep` to absolute paths for secure execution. `isPathSafe()` validates PATH environment variable composition. Satisfies SonarCloud security hotspot requirements for `execFileSync`/`execSync` PATH safety.
 
 ### Fixed
 
+- **SonarCloud security hotspots — 15 resolved** — Addressed all flagged items:
+  - **11 ReDoS-prone regex patterns**: Replaced greedy `.{0,50}` in fetcher's content quality check with lazy quantifier `.{0,50}?`; replaced alternation-heavy split regex in bing-copilot with `[^\S\n]*` horizontal whitespace; replaced `[\s\S]*` JSON extraction patterns in synthesis.mjs with `indexOf`/`lastIndexOf` brace matching; replaced `.+?\.` in selectors with `[^.]+`; replaced `\s+\S*$` trim patterns in sources.mjs, common.mjs, and content.mjs with `lastIndexOf` word-boundary detection; replaced markdown link regex in common.mjs with O(n) indexOf-based parser.
+  - **4 PATH-injection hotspots in browser-lifecycle.mjs and chrome.mjs**: Created `resolveSystemCmd()` utility returning absolute paths for `powershell.exe`, `netstat.exe`, `taskkill.exe` (Windows) and `/usr/bin/ps`, `/usr/bin/lsof`, `/usr/sbin/ss`, `/usr/bin/grep` (Unix). Replaced all bare command names in `execFileSync`/`execSync` calls.
+
 - **SonarCloud minor vulnerability false positives** — Confirmed both remaining issues are false positives (internal diagnostic logging in `bin/gschrome.mjs` and test debug output in `test/fetcher-cli.mjs`). Verified via full smoke test suite: all 33 unit tests pass, all 4 engines (Perplexity, Bing, Google, Gemini) return results at all depths (fast/standard/deep), CDP safety wrappers correctly enforce mode boundaries.
 
 - **SonarCloud security hotspots** (re-verified) — All previously fixed hotspots remain resolved: replaced `spawn("node", ...)` with `spawn(process.execPath, ...)`, replaced `Math.random()` with `crypto.randomInt()`, 19 remaining hotspots confirmed as false positives (hardcoded `execSync` commands, simple regex patterns).

package/README.md

@@ -71,9 +71,10 @@ greedy_search({ query: "Bing captcha setup", engine: "bing", visible: true });
 Inside Pi, prefer the extension commands (no package path needed):
 
 ```text
-/greedy-visible  # launch visible Chrome for captcha/login/cookie setup
-/greedy-status   # show GreedySearch Chrome status
-/greedy-kill     # stop GreedySearch Chrome
+/greedy-visible      # launch visible Chrome for captcha/login/cookie setup
+/greedy-status       # show GreedySearch Chrome status
+/greedy-kill         # stop GreedySearch Chrome
+/set-greedy-locale   # set default result language (de, fr, es, ja, etc.)
 ```
 
 Git install path:

package/extractors/bing-aria.mjs

@@ -0,0 +1,539 @@
+#!/usr/bin/env node
+
+// extractors/bing-aria.mjs — ARIA-tree-based Bing Copilot extractor
+//
+// Instead of copy button → clipboard polling → DOM fallback → iframe spelunking,
+// this extractor builds an ARIA accessibility tree of the page, finds the
+// Copilot answer region, and extracts structured text + sources directly.
+//
+// Inspiration: browser-use-rs extract_dom.js (Playwright's ariaSnapshot)
+//
+// Usage:
+//   node extractors/bing-aria.mjs "<query>" [--tab <prefix>]
+//
+// Output (stdout): JSON { answer, sources, query, url }
+// Errors to stderr only — stdout is always clean JSON for piping.
+
+import {
+	cdp,
+	formatAnswer,
+	getOrOpenTab,
+	handleError,
+	injectClipboardInterceptor,
+	jitter,
+	outputJson,
+	parseArgs,
+	parseSourcesFromMarkdown,
+	prepareArgs,
+	TIMING,
+	validateQuery,
+	waitForSelector,
+	waitForStreamComplete,
+} from "./common.mjs";
+import { dismissConsent, handleVerification } from "./consent.mjs";
+import { SELECTORS } from "./selectors.mjs";
+
+const S = SELECTORS.bing;
+
+// ============================================================================
+// ARIA-tree answer extraction
+// ============================================================================
+
+const EXTRACT_ARIA_JS = String.raw`
+(async function() {
+	'use strict';
+
+	// ── visibility helpers ──────────────────────────
+	function isHidden(el) {
+		if (['STYLE','SCRIPT','NOSCRIPT','TEMPLATE'].includes(el.tagName)) return true;
+		const s = window.getComputedStyle(el);
+		if (s.visibility !== 'visible') return true;
+		if (s.display === 'none') return true;
+		if (el.getAttribute('aria-hidden') === 'true') return true;
+		return false;
+	}
+
+	function isVisible(el) {
+		const r = el.getBoundingClientRect();
+		return r.width > 0 && r.height > 0;
+	}
+
+	function getRole(el) {
+		const explicit = el.getAttribute('role');
+		if (explicit) return explicit.split(' ')[0];
+		const tag = el.tagName;
+		const map = {
+			BUTTON:'button', A: el.hasAttribute('href')?'link':null,
+			INPUT: (()=>{ const t=(el.type||'text').toLowerCase(); return {button:'button',checkbox:'checkbox',radio:'radio',range:'slider',search:'searchbox',text:'textbox',email:'textbox',tel:'textbox',url:'textbox',number:'spinbutton'}[t]||'textbox'; })(),
+			TEXTAREA:'textbox', SELECT: el.hasAttribute('multiple')||el.size>1?'listbox':'combobox',
+			H1:'heading',H2:'heading',H3:'heading',H4:'heading',H5:'heading',H6:'heading',
+			IMG: el.getAttribute('alt')===''?'presentation':'img',
+			NAV:'navigation', MAIN:'main', ARTICLE:'article',
+			HEADER:'banner', FOOTER:'contentinfo', ASIDE:'complementary',
+			FORM:'form', TABLE:'table', UL:'list', OL:'list', LI:'listitem',
+			P:'paragraph', DIALOG:'dialog', IFRAME:'iframe'
+		};
+		return map[tag] || 'generic';
+	}
+
+	function getName(el) {
+		const label = el.getAttribute('aria-label');
+		if (label) return label;
+		const labelledBy = el.getAttribute('aria-labelledby');
+		if (labelledBy) {
+			const texts = labelledBy.split(/\s+/).map(id => {
+				const e = document.getElementById(id);
+				return e ? e.textContent : '';
+			}).filter(Boolean);
+			if (texts.length) return texts.join(' ');
+		}
+		if (['INPUT','TEXTAREA','SELECT'].includes(el.tagName)) {
+			const id = el.id;
+			if (id) {
+				const lbl = document.querySelector('label[for="'+id+'"]');
+				if (lbl) return lbl.textContent || '';
+			}
+			const parentLbl = el.closest('label');
+			if (parentLbl) return parentLbl.textContent || '';
+			const ph = el.getAttribute('placeholder');
+			if (ph) return ph;
+		}
+		if (el.tagName === 'IMG') return el.getAttribute('alt') || '';
+		const title = el.getAttribute('title');
+		if (title) return title;
+		return '';
+	}
+
+	// ── Build ARIA tree ──────────────────────────
+	let indexCounter = 0;
+
+	function buildTree(node, visited = new Set()) {
+		if (visited.has(node)) return null;
+		visited.add(node);
+
+		if (node.nodeType === 3) { // text node
+			return node.nodeValue;
+		}
+		if (node.nodeType !== 1) return null;
+
+		const el = node;
+		if (isHidden(el)) return null;
+
+		const role = getRole(el);
+		if (!role || role === 'presentation' || role === 'none') return null;
+
+		const name = (getName(el) || '').replace(/\s+/g, ' ').trim();
+		const box = el.getBoundingClientRect();
+		const visible = isVisible(el);
+		const cursor = window.getComputedStyle(el).cursor;
+
+		const result = {
+			role,
+			name,
+			children: [],
+			visible,
+			cursor,
+			tag: el.tagName,
+		};
+
+		// index visible interactive + pointer-cursor elements
+		if (visible && (cursor === 'pointer' || ['button','link','textbox','searchbox',
+			'checkbox','radio','combobox','listbox','option','menuitem',
+			'slider','spinbutton','switch','tab','heading'].includes(role))) {
+			result.index = indexCounter++;
+			if (el.tagName === 'A' && el.href) result.href = el.href;
+			if (el.id) result.id = el.id;
+		}
+
+		// shadow DOM
+		if (el.shadowRoot) {
+			for (let c = el.shadowRoot.firstChild; c; c = c.nextSibling) {
+				const child = buildTree(c, visited);
+				if (child) result.children.push(child);
+			}
+		}
+
+		// regular children
+		for (let c = el.firstChild; c; c = c.nextSibling) {
+			if (c.assignedSlot) continue;
+			const child = buildTree(c, visited);
+			if (child) result.children.push(child);
+		}
+
+		// aria-owns
+		if (el.hasAttribute('aria-owns')) {
+			for (const id of el.getAttribute('aria-owns').split(/\s+/)) {
+				const owned = document.getElementById(id);
+				if (owned) {
+					const child = buildTree(owned, visited);
+					if (child) result.children.push(child);
+				}
+			}
+		}
+
+		return result;
+	}
+
+	// ── Find answer region ──────────────────────────
+	// Locale-agnostic: finds the LAST Copilot AI message container.
+	// Copilot uses consistent CSS patterns across locales:
+	//   - AI messages: class contains "ai-message" or "response"
+	//   - User messages: different class prefix
+	//   - We take the LAST ai-message container in DOM order.
+	function findAnswerRegion() {
+		// Look for AI message containers with the known Copilot class pattern
+		// Tailwind-based: group/ai-message, .response-content, etc.
+		const allDivs = document.querySelectorAll('div[class*="ai-message"], div[class*="response-content"], div[class*="message"]');
+		let best = null;
+		for (const el of allDivs) {
+			const text = (el.innerText || '');
+			if (text.length > 100) best = el;  // take last one with substantial text
+		}
+		if (best) return best;
+
+		// Fallback: walk the DOM looking for containers with role=region/article
+		// and substantial text, take the last one.
+		const containers = [];
+		(function walk(el) {
+			if (!el || el.nodeType !== 1) return;
+			const role = el.getAttribute('role');
+			const cls = (el.className || '').toString();
+			if (role === 'region' || role === 'article' ||
+				cls.includes('ac-container')) {
+				const text = (el.innerText || '');
+				if (text.length > 100) containers.push(el);
+			}
+			for (const c of el.children) walk(c);
+		})(document.body);
+		if (containers.length > 0) return containers[containers.length - 1];
+
+		return document.body;
+	}
+
+	// ── Extract text from tree ──────────────────────
+	// Returns a normalized string. Block elements get newlines.
+	// Inline elements flow with whitespace separation.
+	// Locale-agnostic: filters by role/structure only, never by text content.
+	function extractText(node, isInline = false) {
+		if (typeof node === 'string') return node;
+		if (!node) return '';
+
+		// Skip UI buttons entirely
+		if (node.role === 'button') return '';
+
+		const parts = [];
+
+		// heading → markdown heading with surrounding newlines (block)
+		if (node.role === 'heading') {
+			const level = parseInt(node.tag?.[1]) || 2;
+			const inner = node.children.map(c => extractText(c)).join('');
+			if (inner.trim()) parts.push('\n' + '#'.repeat(level) + ' ' + inner.trim() + '\n');
+		}
+
+		// link → markdown link (inline if inside text, block otherwise)
+		else if (node.role === 'link' && node.href) {
+			const text = node.children.map(c => extractText(c)).join('').trim();
+			if (text && node.href.startsWith('http')) {
+				parts.push('[' + text + '](' + node.href + ')');
+			} else if (text) {
+				parts.push(text);
+			}
+		}
+
+		// listitem → preserve structure (block)
+		else if (node.role === 'listitem') {
+			const text = node.children.map(c => extractText(c)).join('').trim();
+			if (text) parts.push('\n- ' + text);
+		}
+
+		// code blocks
+		else if (node.tag === 'CODE' || node.tag === 'PRE') {
+			const text = node.children.map(c => extractText(c)).join('').trim();
+			if (text) parts.push('\n\x60\x60\x60\n' + text + '\n\x60\x60\x60\n');
+		}
+
+		// paragraph — block level, newlines around
+		else if (node.role === 'paragraph') {
+			const text = node.children.map(c => extractText(c)).join('').trim();
+			if (text) parts.push('\n' + text + '\n');
+		}
+
+		// generic/inline — flow text, join tight (whitespace already in text nodes)
+		else {
+			for (const child of node.children) {
+				parts.push(extractText(child));
+			}
+		}
+
+		return parts.join('');
+	}
+
+	// ── Collect sources ──────────────────────────
+	function collectLinks(node) {
+		const links = [];
+		function walk(n) {
+			if (typeof n === 'string') return;
+			if (!n) return;
+			if (n.role === 'link' && n.href && n.href.startsWith('http') &&
+				!n.href.includes('copilot.microsoft.com') &&
+				!n.href.includes('bing.com') &&
+				!n.href.includes('microsoft.com/privacy')) {
+				links.push({ title: n.name || '', url: n.href });
+			}
+			for (const c of n.children) walk(c);
+		}
+		walk(node);
+		// deduplicate by url
+		const seen = new Set();
+		return links.filter(l => { if (seen.has(l.url)) return false; seen.add(l.url); return true; });
+	}
+
+	// ── Execute ──────────────────────────────────
+	try {
+		// Wait for the answer to actually render — the stream may be "complete"
+		// but React hasn't painted the AI message yet. Poll for ai-message content.
+		await new Promise(r => setTimeout(r, 400));
+
+		const deadline = Date.now() + 8000;
+		let answerEl = null;
+		while (Date.now() < deadline) {
+			answerEl = findAnswerRegion();
+			if (answerEl && (answerEl.innerText || '').length > 200) break;
+			answerEl = null;
+			await new Promise(r => setTimeout(r, 500));
+		}
+
+		if (!answerEl) {
+			return JSON.stringify({ error: 'No answer region found (content too short or not rendered)', answer: '', sources: [] });
+		}
+
+		if (!answerEl) {
+			return JSON.stringify({ error: 'No answer region found', answer: '', sources: [] });
+		}
+
+		const tree = buildTree(answerEl);
+		if (!tree) {
+			return JSON.stringify({ error: 'ARIA tree build failed', answer: '', sources: [] });
+		}
+
+		const text = extractText(tree);
+
+		// Post-process: structural normalization only (locale-agnostic)
+		// Buttons are already filtered by role in extractText.
+		// Deduplication handles Copilot's mobile+desktop DOM variants.
+		let clean = text
+			.replace(/\n{3,}/g, '\n\n')
+			.replace(/^\s+|\s+$/g, '')
+			.trim();
+
+		// Strip leading heading if it's the Copilot "X said" label
+		// (locale-agnostic: just checks for markdown heading syntax at start)
+		clean = clean.replace(/^#{1,6}\s+.+?\n\n/, '');
+
+		// Deduplicate: Copilot sends duplicate DOM for responsive variants
+		const lines = clean.split('\n');
+		const seen = new Set();
+		const deduped = [];
+		for (const line of lines) {
+			const normalized = line.trim();
+			if (!normalized) { deduped.push(''); continue; }
+			if (normalized.length <= 2 && /^[-–—•·]$/.test(normalized)) continue;
+			if (seen.has(normalized)) continue;
+			seen.add(normalized);
+			deduped.push(line);
+		}
+		clean = deduped.join('\n').replace(/\n{3,}/g, '\n\n').trim();
+
+		const sources = collectLinks(tree).slice(0, 10);
+
+		return JSON.stringify({ answer: clean, sources });
+	} catch (e) {
+		return JSON.stringify({ error: e.toString(), answer: '', sources: [] });
+	}
+})()
+`;
+
+async function extractAnswer(tab) {
+	console.error("[bing-aria] Extracting answer via ARIA tree...");
+
+	const resultRaw = await cdp(["eval", tab, EXTRACT_ARIA_JS], 45000);
+
+	let result;
+	try {
+		result = JSON.parse(resultRaw);
+	} catch {
+		throw new Error(
+			`ARIA extraction returned invalid JSON: ${resultRaw.slice(0, 200)}`,
+		);
+	}
+
+	if (result.error) {
+		throw new Error(`ARIA extraction failed: ${result.error}`);
+	}
+
+	const { answer, sources: ariaSources } = result;
+
+	if (!answer || answer.length < 10) {
+		throw new Error(
+			`ARIA extraction returned insufficient content (${answer?.length || 0} chars)`,
+		);
+	}
+
+	// Hybrid: click copy button for markdown sources only (answer already extracted via ARIA).
+	// At this point the copy button is guaranteed rendered — no hydration race, no retries.
+	const GLOBAL_VAR = "__bingAriaClipboard";
+	await injectClipboardInterceptor(tab, GLOBAL_VAR);
+	const clipSources = await grabClipboardSources(tab, GLOBAL_VAR);
+	console.error(`[bing-aria] Clipboard sources: ${clipSources.length}`);
+
+	// Merge: ARIA DOM sources + clipboard markdown sources
+	const allSources = [...ariaSources, ...clipSources]
+		.filter((v, i, arr) => arr.findIndex((x) => x.url === v.url) === i)
+		.slice(0, 10);
+
+	console.error(
+		`[bing-aria] Extracted ${answer.length} chars, ${allSources.length} sources`,
+	);
+	return { answer: answer.trim(), sources: allSources };
+}
+
+/**
+ * Click the Bing copy button, grab clipboard markdown, extract just the [title](url) sources.
+ * Lightweight — no retries, no hydration delay (button is already visible at this point).
+ */
+async function grabClipboardSources(tab, globalVar) {
+	try {
+		// Click last copy button (the most recent AI message)
+		const copyBtn = S.copyButton;
+		await cdp([
+			"eval",
+			tab,
+			`(() => {
+				window.${globalVar} = '';
+				const buttons = document.querySelectorAll('${copyBtn}');
+				buttons[buttons.length - 1]?.click();
+			})()`,
+		]);
+
+		// Poll clipboard briefly (2s max — if it doesn't work, no big deal)
+		const deadline = Date.now() + 2000;
+		while (Date.now() < deadline) {
+			const text = await cdp(["eval", tab, `window.${globalVar} || ''`]).catch(
+				() => "",
+			);
+			if (text && text.length > 20) {
+				return parseSourcesFromMarkdown(text);
+			}
+			await new Promise((r) => setTimeout(r, 200));
+		}
+	} catch (e) {
+		console.error(`[bing-aria] Clipboard source grab failed: ${e.message}`);
+	}
+	return [];
+}
+
+// ============================================================================
+// Main
+// ============================================================================
+
+const USAGE =
+	'Usage: node extractors/bing-aria.mjs "<query>" [--tab <prefix>]\n';
+
+async function main() {
+	const args = await prepareArgs(process.argv.slice(2));
+	validateQuery(args, USAGE);
+
+	const { query, tabPrefix, short } = parseArgs(args);
+
+	try {
+		if (!tabPrefix) await cdp(["list"]);
+		const tab = await getOrOpenTab(tabPrefix);
+
+		const currentUrl = await cdp(["eval", tab, "document.location.href"]).catch(
+			() => "",
+		);
+		let onCopilot = false;
+		try {
+			const host = new URL(currentUrl).hostname.toLowerCase();
+			onCopilot =
+				host === "copilot.microsoft.com" ||
+				host.endsWith(".copilot.microsoft.com");
+		} catch {}
+
+		if (!onCopilot) {
+			await cdp(["nav", tab, "https://copilot.microsoft.com/"], 20000);
+			await new Promise((r) => setTimeout(r, 600));
+		}
+		await dismissConsent(tab, cdp);
+
+		const verifyResult = await handleVerification(tab, cdp, 10000);
+		if (verifyResult === "needs-human") {
+			throw new Error(
+				"Copilot verification required — please solve it manually in the browser window",
+			);
+		}
+
+		if (verifyResult === "clicked") {
+			await new Promise((r) => setTimeout(r, TIMING.afterVerify));
+			const currentUrl = await cdp([
+				"eval",
+				tab,
+				"document.location.href",
+			]).catch(() => "");
+			let onCopilot = false;
+			try {
+				const host = new URL(currentUrl).hostname.toLowerCase();
+				onCopilot =
+					host === "copilot.microsoft.com" ||
+					host.endsWith(".copilot.microsoft.com");
+			} catch {}
+			if (!onCopilot) {
+				await cdp(["nav", tab, "https://copilot.microsoft.com/"], 20000);
+				await new Promise((r) => setTimeout(r, 600));
+				await dismissConsent(tab, cdp);
+			}
+		}
+
+		const inputReady = await waitForSelector(tab, S.input, 15000, 500);
+		await new Promise((r) => setTimeout(r, jitter(300)));
+
+		if (!inputReady) {
+			throw new Error(
+				"Copilot input not found — verification may have failed or page is in unexpected state",
+			);
+		}
+
+		// NO clipboard interceptor needed — ARIA extraction reads the DOM directly
+		await cdp(["click", tab, S.input]);
+		await new Promise((r) => setTimeout(r, TIMING.postClick));
+		await cdp(["type", tab, query]);
+		await new Promise((r) => setTimeout(r, TIMING.postType));
+
+		await cdp([
+			"eval",
+			tab,
+			`document.querySelector('${S.input}')?.dispatchEvent(new KeyboardEvent('keydown',{key:'Enter',bubbles:true,keyCode:13})), 'ok'`,
+		]);
+
+		// Wait for Copilot's response to finish streaming
+		await waitForStreamComplete(tab, { timeout: 60000, minLength: 50 });
+
+		const { answer, sources } = await extractAnswer(tab);
+		if (!answer)
+			throw new Error("No answer extracted — Copilot may not have responded");
+
+		const finalUrl = await cdp(["eval", tab, "document.location.href"]).catch(
+			() => "",
+		);
+		outputJson({
+			query,
+			url: finalUrl,
+			answer: formatAnswer(answer, short),
+			sources,
+		});
+	} catch (e) {
+		handleError(e);
+	}
+}
+
+main();

package/extractors/bing-copilot.mjs

@@ -115,7 +115,7 @@ async function extractFromVisibleDom(tab) {
 			.split(/Copilot said\s*/i)
 			.pop()
 			.split(
-				/\n\s*(?:Good response|Bad response|Share message|Copy message|Read aloud|Regenerate|Edit in a page|Message Copilot|Smart)\b/i,
+				/\n[^\S\n]*(?:Good response|Bad response|Share message|Copy message|Read aloud|Regenerate|Edit in a page|Message Copilot|Smart)(?![\w])/i,
 			)[0]
 			.trim();