\n View your index\n
\n\n\n\n\n\n\n Notice a bug? Open an issue!\n
\n\n" }, { "path": "public/problem_find.mjs", "content": "#!/usr/bin/env node\n\nimport fs from 'node:fs';\nimport path from 'node:path';\nimport child_process from 'node:child_process';\n\nimport {\n loadPref,\n cache_file,\n index_file,\n} from '../src/args.js';\n\nconst CLEAN = false;\nconst CONCURRENT = 7;\nconst sleep = ms => new Promise(res => setTimeout(res, ms));\nconst problems = new Map();\nlet cleaning = false;\nlet made = false;\n\nprocess.on('exit', cleanup);\nprocess.on('SIGINT', cleanup);\nprocess.on('SIGTERM', cleanup);\nprocess.on('SIGHUP', cleanup);\nprocess.on('SIGUSR2', cleanup);\nprocess.on('beforeExit', cleanup);\n\nconsole.log({Pref:loadPref(), cache_file: cache_file(), index_file: index_file()});\nmake();\n\nasync function make() {\n const indexFile = fs.readFileSync(index_file()).toString();\n JSON.parse(indexFile).map(([key, value]) => {\n if ( typeof key === \"number\" ) return;\n if ( key.startsWith('ndx') ) return;\n if ( value.title === undefined ) {\n console.log('no title property', {key, value});\n }\n const url = key;\n const title = value.title.toLocaleLowerCase();\n if ( title.length === 0 || title.includes('404') || title.includes('not found') ) {\n if ( problems.has(url) ) {\n console.log('Found duplicate', url, title, problems.get(url));\n }\n problems.set(url, title);\n }\n });\n\n made = true;\n\n cleanup();\n}\n\nfunction cleanup() {\n if ( cleaning ) return;\n if ( ! made ) return;\n cleaning = true;\n console.log('cleanup running');\n const outData = [...problems.entries()];\n fs.writeFileSync(\n path.resolve('.', 'url-problems.json'), \n JSON.stringify(outData, null, 2)\n );\n const {size:bytesWritten} = fs.statSync(\n path.resolve('.', 'url-problems.json'), \n {bigint: true}\n );\n console.log(`Wrote ${outData.length} problem urls in ${bytesWritten} bytes.`);\n process.exit(0);\n}\n\nfunction clean(urlString) {\n const url = new URL(urlString);\n if ( url.hash.startsWith('#!') || url.hostname.includes('google.com') || url.hostname.includes('80s.nyc') ) {\n } else {\n url.hash = '';\n }\n for ( const [key, value] of url.searchParams ) {\n if ( key.startsWith('utm_') ) {\n url.searchParams.delete(key);\n }\n }\n url.pathname = url.pathname.replace(/\\/$/, '');\n url.protocol = 'https:';\n url.pathname = url.pathname.replace(/(\\.htm.?|\\.php)$/, '');\n if ( url.hostname.startsWith('www.') ) {\n url.hostname = url.hostname.replace(/^www./, '');\n }\n const key = url.toString();\n return key;\n}\n\nfunction clean2(urlString) {\n const url = new URL(urlString);\n url.pathname = ''; \n return url.toString();\n}\n\nfunction curlCommand(url) {\n return `curl -k -L -s -o /dev/null -w '%{url_effective}' ${JSON.stringify(url)} \\\n -H 'Accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.9' \\\n -H 'Accept-Language: en,en-US;q=0.9,zh-TW;q=0.8,zh-CN;q=0.7,zh;q=0.6,ja;q=0.5' \\\n -H 'Cache-Control: no-cache' \\\n -H 'Connection: keep-alive' \\\n -H 'DNT: 1' \\\n -H 'Pragma: no-cache' \\\n -H 'Sec-Fetch-Dest: document' \\\n -H 'Sec-Fetch-Mode: navigate' \\\n -H 'Sec-Fetch-Site: none' \\\n -H 'Sec-Fetch-User: ?1' \\\n -H 'Upgrade-Insecure-Requests: 1' \\\n -H 'User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/104.0.0.0 Safari/537.36' \\\n -H 'sec-ch-ua: \"Chromium\";v=\"104\", \" Not A;Brand\";v=\"99\", \"Google Chrome\";v=\"104\"' \\\n -H 'sec-ch-ua-mobile: ?0' \\\n -H 'sec-ch-ua-platform: \"macOS\"' \\\n --compressed ;\n `;\n}\n" }, { "path": "public/redirector.html", "content": "\n\nThis navigation was prevented by 22120 as a Chrome bug fix for some requests causing issues.
\n`).toString(\"base64\");\nexport const BLOCKED_HEADERS = [\n {name: \"X-Powered-By\", value: \"Dosyago-Corporation\"},\n {name: \"X-Blocked-Internally\", value: \"Custom 22120 Chrome bug fix\"},\n {name: \"Accept-Ranges\", value: \"bytes\"},\n {name: \"Cache-Control\", value: \"public, max-age=0\"},\n {name: \"Content-Type\", value: \"text/html; charset=UTF-8\"},\n {name: \"Content-Length\", value: `${BLOCKED_BODY.length}`}\n];\n\nconst BLOCKED_RESPONSE = `\nHTTP/1.1 ${BLOCKED_CODE} OK\nX-Powered-By: Zanj-Dosyago-Corporation\nX-Blocked-Internally: Custom ad blocking\nAccept-Ranges: bytes\nCache-Control: public, max-age=0\nContent-Type: text/html; charset=UTF-8\nContent-Length: ${BLOCKED_BODY.length}\n\n${BLOCKED_BODY}\n`;\n\nexport default BLOCKED_RESPONSE;\n\n" }, { "path": "src/bookmarker.js", "content": "import os from 'os';\nimport Path from 'path';\nimport fs from 'fs';\n\nimport {DEBUG as debug} from './common.js';\n\nconst DEBUG = debug || false;\n// Chrome user data directories by platform. \n // Source 1: https://chromium.googlesource.com/chromium/src/+/HEAD/docs/user_data_dir.md \n // Source 2: https://superuser.com/questions/329112/where-are-the-user-profile-directories-of-google-chrome-located-in\n\nconst FS_WATCH_OPTS = {\n persistent: false,\n};\n\n// Note:\n // Not all the below are now used or supported by this code\nconst UDD_PATHS = {\n 'win': '%LOCALAPPDATA%\\\\Google\\\\Chrome\\\\User Data',\n 'winxp' : '%USERPROFILE%\\\\Local Settings\\\\Application Data\\\\Google\\\\Chrome\\\\User Data',\n 'macos' : Path.resolve(os.homedir(), 'Library/Application Support/Google/Chrome'),\n 'nix' : Path.resolve(os.homedir(), '.config/google-chrome'),\n 'chromeos': '/home/chronos', /* no support */\n 'ios': 'Library/Application Support/Google/Chrome', /* no support */\n};\nconst PLAT_TABLE = {\n 'darwin': 'macos',\n 'linux': 'nix'\n};\nconst PROFILE_DIR_NAME_REGEX = /^(Default|Profile \\d+)$/i;\nconst isProfileDir = name => PROFILE_DIR_NAME_REGEX.test(name);\nconst BOOKMARK_FILE_NAME_REGEX = /^Bookmarks$/i;\nconst isBookmarkFile = name => BOOKMARK_FILE_NAME_REGEX.test(name);\nconst State = {\n active: new Set(), /* active Bookmark files (we don't know these until file changes) */\n books: {\n\n }\n};\n\nexport async function* bookmarkChanges() {\n // try to get the profile directory\n const rootDir = getProfileRootDir();\n\n if ( !fs.existsSync(rootDir) ) {\n throw new TypeError(`Sorry! The directory where we thought the Chrome profile directories may be found (${rootDir}), does not exist. We can't monitor changes to your bookmarks, so Bookmark Select Mode is not supported.`);\n }\n\n // state constants and variables (including chokidar file glob observer)\n const observers = [];\n const ps = [];\n let change = false;\n let notifyChange = false;\n let stopLooping = false;\n let shuttingDown = false;\n\n // create sufficient observers\n const dirs = fs.readdirSync(rootDir, {withFileTypes:true}).reduce((Files, dirent) => {\n if ( dirent.isDirectory() && isProfileDir(dirent.name) ) {\n const filePath = Path.resolve(rootDir, dirent.name);\n\n if ( fs.existsSync(filePath) ) {\n Files.push(filePath); \n }\n }\n return Files;\n }, []);\n for( const dirPath of dirs ) {\n // first read it in\n const filePath = Path.resolve(dirPath, 'Bookmarks');\n if ( fs.existsSync(filePath) ) {\n const data = fs.readFileSync(filePath);\n const jData = JSON.parse(data);\n State.books[filePath] = flatten(jData, {toMap:true});\n }\n\n const observer = fs.watch(dirPath, FS_WATCH_OPTS);\n console.log(`Observing ${dirPath}`);\n // Note\n // allow the parent process to exit \n //even if observer is still active somehow\n observer.unref();\n\n // listen for all events from the observer\n observer.on('change', (event, filename) => {\n filename = filename || '';\n // listen to everything\n const path = Path.resolve(dirPath, filename);\n DEBUG.verboseSlow && console.log(event, path);\n if ( isBookmarkFile(filename) ) {\n if ( ! State.active.has(path) ) {\n State.active.add(path);\n }\n // but only act if it is a bookmark file\n DEBUG.verboseSlow && console.log(event, path, notifyChange);\n // save the event type and file it happened to\n change = {event, path};\n // drop the most recently pushed promise from our bookkeeping list\n ps.pop();\n // resolve the promise in the wait loop to process the bookmark file and emit the changes\n notifyChange && notifyChange();\n }\n });\n observer.on('error', error => {\n console.warn(`Bookmark file observer for ${dirPath} error`, error);\n observers.slice(observers.indexOf(observer), 1);\n if ( observers.length ) {\n notifyChange && notifyChange();\n } else {\n stopLooping && stopLooping();\n }\n });\n observer.on('close', () => {\n console.info(`Observer for ${dirPath} closed`);\n observers.slice(observers.indexOf(observer), 1);\n if ( observers.length ) {\n notifyChange && notifyChange();\n } else {\n stopLooping && stopLooping();\n }\n });\n\n observers.push(observer);\n }\n\n // make sure we kill the watcher on process restart or shutdown\n process.on('SIGTERM', shutdown);\n process.on('SIGHUP', shutdown);\n process.on('SIGINT', shutdown);\n process.on('SIGBRK', shutdown);\n\n // the main wait loop that enables us to turn a traditional NodeJS eventemitter\n // into an asychronous stream generator\n waiting: while(true) {\n // Note: code resilience\n //the below two statements can come in any order in this loop, both work\n\n // get, process and publish changes\n // only do if the change is there (first time it won't be because\n // we haven't yielded out (async or yield) yet)\n if ( change ) {\n const {path} = change;\n change = false;\n\n try {\n const changes = flatten(\n JSON.parse(fs.readFileSync(path)), \n {toMap:true, map: State.books[path]}\n );\n\n for( const changeEvent of changes ) yield changeEvent;\n } catch(e) {\n console.warn(`Error publishing Bookmarks changes`, e);\n }\n }\n\n // wait for the next change\n // always wait tho (to allow queueing of the next event to process)\n try {\n await new Promise((res, rej) => {\n // save these\n notifyChange = res; // so we can turn the next turn of the loop\n stopLooping = rej; // so we can break out of the loop (on shutdown)\n ps.push({res,rej}); // so we can clean up any left over promises\n });\n } catch { \n ps.pop();\n break waiting; \n }\n }\n\n shutdown();\n\n return true;\n\n async function shutdown() {\n if ( shuttingDown ) return;\n shuttingDown = true;\n console.log('Bookmark observer shutting down...');\n // clean up any outstanding waiting promises\n while ( ps.length ) {\n /* eslint-disable no-empty */\n try { ps.pop().rej(); } finally {}\n /* eslint-enable no-empty */\n }\n // stop the waiting loop\n stopLooping && setTimeout(() => stopLooping('bookmark watching stopped'), 0);\n // clean up any observers\n while(observers.length) {\n /* eslint-disable no-empty */\n try { observers.pop().close(); } finally {}\n /* eslint-enable no-empty */\n }\n console.log('Bookmark observer shut down cleanly.');\n }\n}\n\nexport function hasBookmark(url) {\n return Object.keys(State.books).filter(key => {\n if ( State.active.size == 0 ) return true; \n return State.active.has(key);\n }).map(key => State.books[key])\n .some(map => map.has(url));\n}\n\nfunction getProfileRootDir() {\n const plat = os.platform();\n let name = PLAT_TABLE[plat];\n let rootDir;\n\n DEBUG.verboseSlow && console.log({plat, name});\n\n if ( !name ) {\n if ( plat === 'win32' ) {\n // because Chrome profile dir location only changes in XP\n // we only care if it's XP or not and so\n // we try to resolve based on the version major and minor (given by release)\n // source: https://docs.microsoft.com/en-us/windows/win32/sysinfo/operating-system-version?redirectedfrom=MSDN\n const rel = os.release();\n const ver = parseFloat(rel); \n if ( !Number.isNaN(ver) && ver <= 5.2 ) {\n // this should be reliable\n name = 'winxp';\n } else {\n // this may not be reliable, but we just do it\n name = 'win';\n }\n } else {\n throw new TypeError(\n `Sorry! We don't know how to find the default Chrome profile on OS platform: ${plat}`\n );\n }\n }\n\n if ( UDD_PATHS[name] ) {\n rootDir = Path.resolve(resolveEnvironmentVariablesToPathSegments(UDD_PATHS[name]));\n } else {\n throw new TypeError(\n `Sorry! We don't know how to find the default Chrome profile on OS name: ${name}`\n );\n }\n\n return rootDir;\n}\n\nfunction flatten(bookmarkObj, {toMap: toMap = false, map} = {}) {\n const nodes = [...Object.values(bookmarkObj.roots)];\n const urls = toMap? (map || new Map()) : [];\n const urlSet = new Set();\n const changes = [];\n\n while(nodes.length) {\n const next = nodes.pop();\n const {name, type, url} = next;\n switch(type) {\n case \"url\":\n if ( toMap ) {\n if ( map ) {\n if ( urls.has(url) ) {\n const {name:oldName} = urls.get(url);\n if ( name !== oldName ) {\n if ( !urlSet.has(url) ) {\n changes.push({\n type: \"Title updated\",\n url,\n oldName, \n name\n });\n }\n }\n } else {\n changes.push({\n type: \"new\",\n name, url\n });\n }\n } \n if ( !urlSet.has(url) ) {\n urls.set(url, next);\n }\n urlSet.add(url);\n } else {\n urls.push(next);\n }\n break;\n case \"folder\":\n nodes.push(...next.children);\n break;\n default:\n console.info(\"New type\", type, next);\n break;\n \n }\n }\n\n if (map) {\n [...map.keys()].forEach(url => {\n if ( !urlSet.has(url) ) {\n changes.push({\n type: \"delete\",\n url\n });\n map.delete(url);\n }\n });\n }\n\n return map ? changes : urls;\n}\n\n// source: https://stackoverflow.com/a/33017068\nfunction resolveEnvironmentVariablesToPathSegments(path) {\n return path.replace(/%([^%]+)%/g, function(_, key) {\n return process.env[key];\n });\n}\n\n/*\ntest();\nasync function test() {\n for await ( const change of bookmarkChanges() ) {\n console.log(change);\n }\n}\n*/\n\n\n/*\nfunction* profileDirectoryEnumerator(maxN = 9999) {\n let index = 0; \n while(index <= maxN) {\n const profileDirName = index ? `Profile ${index}` : `Default`;\n yield profileDirName;\n }\n}\n*/\n" }, { "path": "src/common.js", "content": "import path from 'path';\nimport {fileURLToPath} from 'url';\nimport fs from 'fs';\nimport os from 'os';\nimport { root } from './root.js';\n\nconst { APP_ROOT: __ROOT } = root;\n\nconst DEEB = process.env.DEBUG_22120_VERBOSE || false;\n\nexport const DEBUG = {\n showBrowser: false,\n verboseBrowser: false,\n showList: false,\n showStatus: false,\n debugSec: false,\n askFirst: true,\n verboseSlow: process.env.VERBOSE_DEBUG_22120 || DEEB,\n debug: process.env.DEBUG_22120 || DEEB,\n verbose: DEEB || process.env.VERBOSE_DEBUG_22120 || process.env.DEBUG_22120,\n checkPred: false,\n}\nexport const SHOW_FETCH = false;\n\nif ( DEBUG.debug ) {\n console.log({__ROOT});\n}\n\n// server related\nexport const PUBLIC_SERVER = true;\n\n// crawl related\nexport const MIN_TIME_PER_PAGE = 10000;\nexport const MAX_TIME_PER_PAGE = 32000;\nexport const MIN_WAIT = 200;\nexport const MAX_WAITS = 300;\nexport const BATCH_SIZE = 5; // crawl batch size (how many concurrent tabs for crawling)\nexport const MAX_REAL_URL_LENGTH = 2**15 - 1;\n\nexport const CHECK_INTERVAL = 400;\nexport const TEXT_NODE = 3;\nexport const MAX_HIGHLIGHTABLE_LENGTH = 0; /* 0 is no max length for highlight */\nexport const MAX_TITLE_LENGTH = 140;\nexport const MAX_URL_LENGTH = 140;\nexport const MAX_HEAD = 140;\n\nconst LOCALP = path.resolve(os.homedir(), 'local-sslcerts', 'privkey.pem');\nconst ANYP = path.resolve(os.homedir(), 'sslcerts', 'privkey.pem');\nexport const GO_SECURE = fs.existsSync(LOCALP) || fs.existsSync(ANYP);\nconst cert_path = GO_SECURE ? path.dirname(fs.existsSync(LOCALP) ? LOCALP : fs.existsSync(ANYP) ? ANYP : null) : null;\nexport const CERT_PATH = () => GO_SECURE ? cert_path : false;\n\nexport class RichError extends Error {\n constructor(msg) {\n super(msg);\n let textMessage;\n try {\n textMessage = JSON.stringify(msg);\n } catch(e) {\n console.warn(`Could not create RichError from argument ${msg.toString ? msg.toString() : msg} as JSON serialization failed. RichError argument MUST be JSON serializable. Failure error was:`, e);\n return;\n }\n super(textMessage);\n }\n}\n\n/* text nodes inside these elements that are ignored */\nexport const FORBIDDEN_TEXT_PARENT = new Set([\n 'STYLE',\n 'SCRIPT',\n 'NOSCRIPT',\n /* we could remove these last two so as to index them as well */\n 'DATALIST',\n 'OPTION'\n]);\nexport const ERROR_CODE_SAFE_TO_IGNORE = new Set([\n -32000, /* message:\n Can only get response body on requests captured after headers received.\n * ignore because: \n seems to only happen when new navigation aborts all \n pending requests of the unloading page \n */\n -32602, /* message:\n Invalid InterceptionId.\n * ignore because: \n seems to only happen when new navigation aborts all \n pending requests of the unloading page \n */\n]);\n\nexport const SNIP_CONTEXT = 31;\n\nexport const NO_SANDBOX = (process.env.DEBUG_22120 && process.env.SET_22120_NO_SANDBOX) || false;\n\nexport const APP_ROOT = __ROOT;\n\nexport const sleep = ms => new Promise(res => setTimeout(res, ms));\n\nexport function say(o) {\n console.log(JSON.stringify(o));\n}\n\nexport function clone(o) {\n return JSON.parse(JSON.stringify(o));\n}\n\nexport async function untilTrue(pred, waitOverride = MIN_WAIT, maxWaits = MAX_WAITS) {\n if ( waitOverride < 0 ) {\n maxWaits = -1;\n waitOverride = MIN_WAIT;\n }\n let waitCount = 0;\n let resolve;\n const pr = new Promise(res => resolve = res);\n setTimeout(checkPred, 0);\n return pr;\n\n async function checkPred() {\n DEBUG.checkPred && console.log('Checking', pred.toString());\n if ( await pred() ) {\n return resolve(true);\n } else {\n waitCount++;\n if ( waitCount < maxWaits || maxWaits < 0 ) {\n setTimeout(checkPred, waitOverride);\n } else {\n resolve(false);\n }\n }\n }\n}\n" }, { "path": "src/gem-highlighter.js", "content": "// highlighter.js\n\nimport ukkonen from 'ukkonen';\nimport {DEBUG} from './common.js';\n\nconst MAX_ACCEPT_SCORE = 0.5;\nconst CHUNK_SIZE = 12; // Default, can be overridden\n\nfunction params(qLength, chunkSize = CHUNK_SIZE) {\n // MaxDist is the maximum edit distance we're willing to consider for a chunk.\n // If chunkSize is small, MaxDist might be too restrictive for longer queries.\n // Consider making MaxDist also a function of qLength, e.g., Math.min(chunkSize, qLength / 2)\n const MaxDist = chunkSize; // This was the original.\n // A more flexible MaxDist could be:\n // const MaxDist = Math.min(chunkSize, Math.floor(qLength * 0.4) + 1); // Allow up to 40% of query length as edits\n\n // MinScore: if distance is 0, this is the \"best\" raw score.\n // If qLength and chunkSize are very different, distance can't be 0.\n // The distance between two strings is at least abs(len1 - len2).\n const MinScore = Math.abs(qLength - chunkSize);\n\n // MaxScore: used for scaling. (distance - MinScore) / MaxScore_Range\n // The maximum possible distance is max(qLength, chunkSize).\n // So, the range of distances is from MinScore to max(qLength, chunkSize).\n // The length of this range is max(qLength, chunkSize) - MinScore.\n const MaxScore_Range = Math.max(qLength, chunkSize) - MinScore;\n\n // If MaxScore_Range is 0 (e.g., qLength === chunkSize, so MinScore is 0, and max distance is qLength),\n // avoid division by zero. In this case, any distance > 0 is \"bad\".\n // A distance of 0 would be a perfect match.\n return {MaxDist, MinScore, MaxScore_Range: MaxScore_Range === 0 ? 1 : MaxScore_Range};\n}\n\n// Helper to wrap query terms with tags within a text\nfunction markText(text, query) {\n if (!text || !query) return text;\n try {\n // Case-insensitive replacement\n const regex = new RegExp('(' + query.replace(/[.*+?^${}()|[\\]\\\\]/g, '\\\\$&') + ')', 'gi');\n return text.replace(regex, '$1');\n } catch (e) {\n // Regex might fail for complex queries, fallback to original text\n console.warn(\"Marking text failed for query:\", query, e);\n return text;\n }\n}\n\nexport function highlight(query, doc, {\n maxLength = 0,\n maxAcceptScore = MAX_ACCEPT_SCORE,\n chunkSize = CHUNK_SIZE,\n // NEW options for server integration\n around = '', // e.g. '' - but we'll handle this internally now\n before = '', // e.g. '' - but we'll handle this internally now\n numResults = 3, // How many top snippets to return\n contextChars = 30 // How many characters before/after the matched chunk\n} = {}) {\n if (chunkSize % 2 !== 0 && chunkSize !== 1) { // Allow chunkSize 1 for exact char matching if desired\n // Original code threw error for odd chunkSize.\n // Relaxing this slightly, but even is generally better for the overlapping strategy.\n // For simplicity, let's stick to the original constraint or make it more robust.\n // For now, let's assume it's usually even or we adjust.\n // If we keep the original overlapping strategy, even chunkSize is important for the offset.\n // Let's keep the original constraint for now:\n if ( chunkSize % 2 ) {\n console.warn(`highlight: chunkSize should ideally be even. Was: ${chunkSize}. Adjusting to ${chunkSize+1}`);\n chunkSize = chunkSize + 1; // Or throw error as original\n }\n }\n\n const originalDocString = doc; // Keep the original string for final snippet extraction\n doc = Array.from(doc); // Work with array of characters for unicode safety\n\n if (maxLength > 0 && doc.length > maxLength) {\n doc = doc.slice(0, maxLength);\n }\n\n if (doc.length === 0 || query.trim() === \"\") {\n return []; // No document or query, no highlights\n }\n\n const queryChars = Array.from(query.toLocaleLowerCase()); // Lowercase query once\n const qLength = queryChars.length;\n\n if (qLength === 0) return [];\n\n\n // --- Fragment Generation ---\n // The original code created two sets of fragments with different offsets.\n // This is a strategy to catch matches that might fall across non-overlapping chunk boundaries.\n // Let's simplify this for clarity first, then consider re-adding if necessary.\n // A simpler approach: overlapping chunks.\n const step = Math.max(1, Math.floor(chunkSize / 2)); // Create overlapping chunks\n const fragments = [];\n for (let i = 0; i <= doc.length - chunkSize; i += step) {\n const fragmentTextChars = doc.slice(i, i + chunkSize);\n fragments.push({\n text: fragmentTextChars.join(''),\n textChars: fragmentTextChars, // Keep char array for lowercase version\n offset: i,\n // symbols: doc // Reference to the full document character array (for context later)\n // This can be memory intensive if doc is huge.\n // We'll use originalDocString and offsets for context.\n });\n }\n // Add last fragment if doc.length is not a multiple of step\n if (doc.length % chunkSize !== 0 && doc.length > chunkSize) {\n const i = Math.floor((doc.length - chunkSize)/step) * step; // last full step\n if (i + chunkSize < doc.length) { // if there's a remainder smaller than chunkSize\n const remainderOffset = i + step > doc.length - chunkSize ? doc.length - chunkSize : i + step;\n if (remainderOffset < doc.length -1 && remainderOffset > 0) { // ensure it's a valid offset\n const fragmentTextChars = doc.slice(remainderOffset, Math.min(remainderOffset + chunkSize, doc.length));\n if (fragmentTextChars.length > 0) {\n fragments.push({\n text: fragmentTextChars.join(''),\n textChars: fragmentTextChars,\n offset: remainderOffset,\n });\n }\n }\n } else if (doc.length < chunkSize) { // if doc is smaller than chunksize\n // The loop for fragments won't run, so add the whole doc as one fragment\n if (fragments.length === 0) {\n fragments.push({\n text: doc.join(''),\n textChars: doc,\n offset: 0,\n });\n }\n }\n }\n if (fragments.length === 0 && doc.length > 0) { // Case: doc is shorter than chunkSize\n fragments.push({\n text: doc.join(''),\n textChars: doc,\n offset: 0,\n });\n }\n\n\n DEBUG.verboseSlow && console.log(\"Generated fragments:\", fragments.length);\n\n const { MaxDist, MinScore, MaxScore_Range } = params(qLength, chunkSize);\n\n const scoredFragments = fragments.map(fragment => {\n const fragmentTextLower = fragment.textChars.join('').toLocaleLowerCase();\n // Ukkonen distance between the lowercase query and lowercase fragment text\n const distance = ukkonen(queryChars.join(''), fragmentTextLower, MaxDist);\n \n // Scale the score: 0 is best (perfect match or close), 1 is worst (MaxDist or more)\n // If distance is -1 (meaning it exceeded MaxDist), assign a very high score.\n let scaledScore;\n if (distance === -1) {\n scaledScore = Infinity; // Or a value > 1, e.g., 2\n } else {\n // scaledScore = (distance - MinScore) / MaxScore_Range;\n // Simpler scaling: distance / qLength (fraction of query that is \"wrong\")\n // This makes maxAcceptScore more intuitive (e.g., 0.2 means up to 20% difference)\n scaledScore = distance / Math.max(1, qLength); // Avoid division by zero for empty query (already handled)\n }\n \n return { score: scaledScore, fragment };\n });\n\n // Sort by score (ascending, lower is better)\n scoredFragments.sort((a, b) => a.score - b.score);\n\n DEBUG.verboseSlow && console.log(\"Top 5 scored fragments:\", scoredFragments.slice(0, 5));\n\n const bestHighlights = [];\n const seenOffsets = new Set(); // To avoid overly similar/overlapping snippets\n\n for (const { score, fragment } of scoredFragments) {\n if (bestHighlights.length >= numResults * 2) break; // Get a slightly larger pool initially\n\n if (score > maxAcceptScore) {\n // If even the best scores are too high, we might not have good matches.\n // However, if we have *some* results already, we might stop.\n // If bestHighlights is empty and score > maxAcceptScore, then we have no good matches.\n if (bestHighlights.length === 0 && score !== Infinity) { // If it's the first one and bad, but not impossible\n // Potentially keep it if we want to *always* return something\n } else if (score === Infinity || score > maxAcceptScore) {\n continue; // Skip clearly bad or too fuzzy matches if we have better options\n }\n }\n \n // Check for overlap with already selected highlights\n let isOverlapping = false;\n for (const existingOffset of seenOffsets) {\n if (Math.abs(fragment.offset - existingOffset) < chunkSize / 2) { // Heuristic for overlap\n isOverlapping = true;\n break;\n }\n }\n if (isOverlapping) continue;\n\n bestHighlights.push({ score, fragment });\n seenOffsets.add(fragment.offset);\n }\n \n DEBUG.verboseSlow && console.log(\"Filtered bestHighlights (before context/marking):\", bestHighlights.length);\n\n\n if (bestHighlights.length === 0 && scoredFragments.length > 0 && scoredFragments[0].score !== Infinity) {\n // If no highlights passed the filter but there was at least one scorable fragment,\n // take the absolute best one, regardless of maxAcceptScore, to ensure we return *something*.\n // This was the behavior of the original code's \"Zero highlights, showing first score\"\n if (scoredFragments[0].fragment) { // Check if fragment exists\n bestHighlights.push(scoredFragments[0]);\n DEBUG.verboseSlow && console.log('No highlights passed filters, taking the absolute best scored fragment.');\n }\n }\n\n\n // Now, construct the final snippets with context and tags\n const finalSnippets = bestHighlights\n .slice(0, numResults) // Take the top N results\n .map(({ score, fragment }) => {\n const start = Math.max(0, fragment.offset - contextChars);\n const end = Math.min(originalDocString.length, fragment.offset + fragment.text.length + contextChars);\n \n let snippetText = originalDocString.substring(start, end);\n\n // Apply tags. This is the crucial part for server integration.\n // We mark the original query within this expanded snippet.\n snippetText = markText(snippetText, query);\n\n return {\n // score, // Optionally include score if useful for UI\n fragment: {\n text: snippetText,\n offset: fragment.offset, // Original offset of the core matched chunk\n // No need for 'symbols' anymore in the returned fragment\n }\n };\n });\n \n // The original code had a \"better\" loop that re-scored with more context.\n // This can be useful but adds complexity. For now, the above provides context around the best chunks.\n // If re-scoring is desired:\n // 1. Take top N initial highlights.\n // 2. For each, expand context (as done above).\n // 3. Re-run Ukkonen on this expanded (but not yet marked) snippet.\n // 4. Re-sort based on these new scores.\n // 5. Then apply tags.\n // This was what your \"better = better.map(hl => { ... })\" loop was doing.\n // Let's defer re-implementing that precisely unless the current results are insufficient.\n\n DEBUG.verboseSlow && console.log(\"Final snippets to return:\", finalSnippets);\n return finalSnippets;\n}\n\n\n// --- trilight function (and its helper getFragmenter) ---\n// This function seems to be an alternative highlighting/segmentation strategy.\n// It's not directly used by the server's current highlight call, but I'll review it.\n\n// (getFragmenter is used by both highlight (implicitly if we restore original frag logic) and trilight)\n// returns a function that creates fragments\nfunction getFragmenter(chunkSize, {overlap = false, step = 1} = {}) {\n if (!Number.isInteger(chunkSize) || chunkSize < 1) {\n throw new TypeError(`chunkSize needs to be a whole number greater than 0`);\n }\n if (!Number.isInteger(step) || step < 1) {\n throw new TypeError(`step needs to be a whole number greater than 0`);\n }\n\n // This function is complex due to its use of reduce and mutating frags array.\n // A generator function or a simple loop might be clearer for fragment generation.\n // However, let's keep its logic for now if it's specific to trilight's needs.\n\n // The original getFragmenter was stateful in a way that's tricky with `reduce`\n // if `overlap` is true and it tries to modify previous elements of `frags`.\n // Let's simplify its signature and usage for `trilight` if it's only for n-grams.\n\n // If for n-grams (overlap=true, step=1 typically for n-grams)\n if (overlap) {\n return function ngramFragmenter(frags, _nextSymbol, index, symbols) {\n if (index <= symbols.length - chunkSize) {\n const ngramChars = symbols.slice(index, index + chunkSize);\n frags.push({\n text: ngramChars.join(''),\n offset: index,\n // symbols: symbols // Avoid if not strictly needed or doc is large\n });\n }\n return frags;\n };\n } else {\n // Non-overlapping chunks (or controlled overlap via step)\n // This is more like the fragment generation now in `highlight`\n return function chunkFragmenter(frags, _nextSymbol, index, symbols) {\n // This will be called for each symbol, which is inefficient for chunking.\n // It's better to do chunking in a loop outside.\n // For now, to match original structure if trilight depends on it:\n if (index % chunkSize === 0) { // Start new chunk\n const chunkChars = symbols.slice(index, Math.min(index + chunkSize, symbols.length));\n if (chunkChars.length > 0) {\n frags.push({\n text: chunkChars.join(''),\n offset: index,\n // symbols: symbols\n });\n }\n }\n return frags;\n };\n }\n}\n\n\nexport function trilight(query, doc, {\n maxLength = 0,\n ngramSize = 3,\n maxSegmentSize = 140,\n numResults = 3 // How many segments to return\n} = {}) {\n const originalDocString = doc; // For final slicing\n query = Array.from(query.toLocaleLowerCase());\n const docCharsLower = Array.from(doc.toLocaleLowerCase());\n \n let effectiveDoc = docCharsLower;\n if (maxLength > 0 && effectiveDoc.length > maxLength) {\n effectiveDoc = effectiveDoc.slice(0, maxLength);\n }\n\n if (effectiveDoc.length === 0 || query.length === 0 || query.length < ngramSize) {\n return [];\n }\n\n // Generate n-grams for document and query\n const docNgrams = [];\n for (let i = 0; i <= effectiveDoc.length - ngramSize; i++) {\n docNgrams.push({ text: effectiveDoc.slice(i, i + ngramSize).join(''), offset: i });\n }\n\n const queryNgrams = [];\n for (let i = 0; i <= query.length - ngramSize; i++) {\n queryNgrams.push({ text: query.slice(i, i + ngramSize).join(''), offset: i });\n }\n \n if (docNgrams.length === 0 || queryNgrams.length === 0) return [];\n\n // Index document n-grams\n const docNgramIndex = new Map();\n for (const ngram of docNgrams) {\n if (!docNgramIndex.has(ngram.text)) {\n docNgramIndex.set(ngram.text, []);\n }\n docNgramIndex.get(ngram.text).push(ngram.offset);\n }\n\n // Find matching n-gram sequences (Longest Common Subsequence of N-gram Offsets)\n // This is essentially what your 'entries' and 'runs' logic is doing.\n // It's finding diagonals in a dot plot of query n-gram index vs doc n-gram index.\n const runs = [];\n for (let qi = 0; qi < queryNgrams.length; qi++) {\n const qNgramText = queryNgrams[qi].text;\n const docOffsets = docNgramIndex.get(qNgramText);\n if (docOffsets) {\n for (const docOffset of docOffsets) {\n // This is a potential start of a run.\n // Try to extend it.\n let currentRunLength = 1;\n let qIdx = qi + 1;\n let dIdx = docOffset + 1; // Next char, not next ngram offset\n // Original logic: dDi = di - lastDi; if (dQi === 1 && dDi === 1)\n // This implies matching characters, not just ngrams.\n // Let's stick to ngram matching for runs.\n // A \"run\" is a sequence of matching n-grams where their relative positions are maintained.\n // q_ngram[i] matches d_ngram[j]\n // q_ngram[i+1] matches d_ngram[j+1] (if ngramSize=1, this is char matching)\n // q_ngram[i+k] matches d_ngram[j+k]\n\n // To find runs more directly:\n // For each match (q_ngram_idx, d_ngram_idx), the value (d_ngram_idx - q_ngram_idx) is constant along a diagonal.\n // Group matches by this diagonal value. Then, within each diagonal, find longest contiguous sequences.\n }\n }\n }\n // The original 'runs' logic is quite specific. Let's try to replicate its intent.\n // It finds consecutive n-grams that match with a consistent offset.\n const entries = [];\n queryNgrams.forEach((qNgram, qNgramIndex) => {\n const docOffsets = docNgramIndex.get(qNgram.text);\n if (docOffsets) {\n docOffsets.forEach(docNgramActualOffset => {\n entries.push({\n qNgramIndex, // Index of the ngram in the query's ngram list\n docNgramActualOffset, // Actual character offset in the document\n text: qNgram.text // The ngram text itself\n });\n });\n }\n });\n\n // Sort entries primarily by document offset, then by query ngram index\n // This helps in identifying consecutive runs.\n entries.sort((a, b) => {\n if (a.docNgramActualOffset !== b.docNgramActualOffset) {\n return a.docNgramActualOffset - b.docNgramActualOffset;\n }\n return a.qNgramIndex - b.qNgramIndex;\n });\n \n const identifiedRuns = [];\n if (entries.length > 0) {\n let currentRun = {\n startDocOffset: entries[0].docNgramActualOffset,\n startQueryNgramIndex: entries[0].qNgramIndex,\n lengthNgrams: 1, // Length in terms of number of ngrams\n // ngrams: [entries[0].text] // For debugging\n };\n\n for (let i = 1; i < entries.length; i++) {\n const prevEntry = entries[i-1];\n const currentEntry = entries[i];\n\n // Check for contiguity:\n // Query ngrams are consecutive: currentEntry.qNgramIndex === prevEntry.qNgramIndex + 1\n // Document ngrams are consecutive (offsets advance by 1 for each char in ngram):\n // currentEntry.docNgramActualOffset === prevEntry.docNgramActualOffset + 1 (if ngrams overlap by n-1)\n // This is the condition from your original code: dQi === 1 && dDi === 1\n // where dDi was char offset difference.\n if (currentEntry.qNgramIndex === (currentRun.startQueryNgramIndex + currentRun.lengthNgrams) &&\n currentEntry.docNgramActualOffset === (currentRun.startDocOffset + currentRun.lengthNgrams) ) {\n currentRun.lengthNgrams++;\n // currentRun.ngrams.push(currentEntry.text);\n } else {\n // End of current run, save it\n identifiedRuns.push({\n docOffset: currentRun.startDocOffset,\n queryNgramStartIndex: currentRun.startQueryNgramIndex,\n // Actual character length of the run in the document:\n // start offset + (num_ngrams - 1) for overlaps + ngramSize for the last one\n docLengthChars: currentRun.lengthNgrams + ngramSize - 1,\n numMatchingNgrams: currentRun.lengthNgrams\n });\n // Start a new run\n currentRun = {\n startDocOffset: currentEntry.docNgramActualOffset,\n startQueryNgramIndex: currentEntry.qNgramIndex,\n lengthNgrams: 1,\n // ngrams: [currentEntry.text]\n };\n }\n }\n // Push the last run\n identifiedRuns.push({\n docOffset: currentRun.startDocOffset,\n queryNgramStartIndex: currentRun.startQueryNgramIndex,\n docLengthChars: currentRun.lengthNgrams + ngramSize - 1,\n numMatchingNgrams: currentRun.lengthNgrams\n });\n }\n \n DEBUG.verboseSlow && console.log(\"Trilight identifiedRuns:\", identifiedRuns);\n\n // The original code then merges runs based on 'gaps'. This is a form of segment clustering.\n // Let's simplify: take the longest runs as primary segments.\n // Sort runs by numMatchingNgrams (as a proxy for quality/length)\n identifiedRuns.sort((a, b) => b.numMatchingNgrams - a.numMatchingNgrams);\n\n const finalSegments = [];\n const addedRunOffsets = new Set();\n\n for (const run of identifiedRuns) {\n if (finalSegments.length >= numResults) break;\n\n // Avoid adding segments that heavily overlap with already chosen ones\n let overlaps = false;\n for(let i = run.docOffset; i < run.docOffset + run.docLengthChars; i++) {\n if (addedRunOffsets.has(i)) {\n overlaps = true;\n break;\n }\n }\n if (overlaps && finalSegments.length > 0) continue; // Allow first segment even if it's the only one\n\n const segmentStart = run.docOffset;\n const segmentEnd = run.docOffset + run.docLengthChars;\n \n // Ensure segment does not exceed maxSegmentSize (original logic was more complex during merging)\n // Here, we just check the individual run. If merging is desired, it's more complex.\n if (run.docLengthChars > maxSegmentSize) {\n // If a single best run is too long, we might truncate it or skip it.\n // For now, let's allow it but be aware.\n // Or, we could try to find a sub-segment centered around the query.\n }\n\n let text = originalDocString.substring(segmentStart, Math.min(segmentEnd, originalDocString.length));\n \n // Mark the text within this segment\n text = markText(text, query.join('')); // query is array of chars\n\n finalSegments.push({\n fragment: { text, offset: segmentStart } // Keep consistent with `highlight` output\n });\n\n for(let i = run.docOffset; i < run.docOffset + run.docLengthChars; i++) {\n addedRunOffsets.add(i);\n }\n }\n \n DEBUG.verboseSlow && console.log(\"Trilight finalSegments:\", finalSegments);\n\n // If no segments, return a small part of the beginning of the document as a fallback\n if (finalSegments.length === 0 && originalDocString.length > 0) {\n DEBUG.verboseSlow && console.log(\"Trilight: No segments found, returning beginning of doc.\");\n let fallbackText = originalDocString.substring(0, Math.min(maxSegmentSize, originalDocString.length));\n fallbackText = markText(fallbackText, query.join(''));\n return [{ fragment: { text: fallbackText, offset: 0 } }];\n }\n\n return finalSegments;\n}\n" }, { "path": "src/hello.js", "content": "console.log(`hello...is it me you're looking for?`);\n" }, { "path": "src/highlighter.js", "content": "// highlighter.js\n\nimport ukkonen from 'ukkonen';\nimport {DEBUG} from './common.js';\n\nconst MAX_ACCEPT_SCORE = 0.5;\nconst CHUNK_SIZE = 12;\n\n// Helper to wrap query terms with tags within a text\n// This function will be used by both highlight and trilight before returning results.\nfunction internalMarkText(textToMark, queryToFind) {\n if (!textToMark || !queryToFind) return textToMark;\n try {\n // Case-insensitive replacement, escaping regex special characters in query\n const escapedQuery = queryToFind.replace(/[.*+?^${}()|[\\]\\\\]/g, '\\\\$&');\n const regex = new RegExp('(' + escapedQuery + ')', 'gi');\n return textToMark.replace(regex, '$1');\n } catch (e) {\n console.warn(\"internalMarkText: Regex failed for query:\", queryToFind, e);\n return textToMark; // Fallback to original text if regex fails\n }\n}\n\n\nfunction calculateUkkonenParams(queryLength, chunkSize = CHUNK_SIZE) {\n // Renamed from 'params' for clarity\n const maxDistance = chunkSize; // Max edit distance for Ukkonen\n const minPossibleScore = Math.abs(queryLength - chunkSize); // Minimum edits based on length difference\n // Max possible score range (denominator for scaling)\n let maxScoreRange = Math.max(queryLength, chunkSize) - minPossibleScore;\n if (maxScoreRange === 0) maxScoreRange = 1; // Avoid division by zero\n\n return {maxDistance, minPossibleScore, maxScoreRange};\n}\n\nexport function highlight(query, docString, {\n maxLength = 0,\n maxAcceptScore = MAX_ACCEPT_SCORE,\n chunkSize = CHUNK_SIZE,\n // Options from server (around, before) are now handled internally by \n // numResults and contextChars are effectively handled by the original logic's\n // \"better.slice(0,3)\" and \"extra\" context respectively.\n} = {}) {\n if (chunkSize % 2) {\n // Original code threw an error. Preserving this behavior.\n throw new TypeError(`chunkSize must be even. Was: ${chunkSize} which is odd.`);\n }\n\n let docChars = Array.from(docString); // Use character array for Unicode safety\n if (maxLength > 0 && docChars.length > maxLength) {\n docChars = docChars.slice(0, maxLength);\n }\n\n if (docChars.length === 0 || query.trim() === \"\") {\n return [];\n }\n\n const queryLower = query.toLocaleLowerCase(); // Lowercase query once\n const queryLength = Array.from(query).length; // Unicode-safe query length\n\n if (queryLength === 0) return [];\n\n const {maxDistance, minPossibleScore, maxScoreRange} = calculateUkkonenParams(queryLength, chunkSize);\n\n // --- Fragment Generation (Identical to original) ---\n // First set of fragments (docChars1)\n const docChars1 = [...docChars]; // Create a mutable copy\n // Pad to make length a multiple of chunkSize\n const padding1Length = (chunkSize - docChars1.length % chunkSize) % chunkSize;\n docChars1.push(...Array(padding1Length).fill(' '));\n const fragments1 = docChars1.reduce(getFragmenter(chunkSize, {symbolsArray: docChars}), []); // Pass original docChars for context\n\n // Second set of fragments (docChars2) with offset\n const docChars2 = [...docChars]; // Create another mutable copy\n // Pad start by half chunkSize\n docChars2.splice(0, 0, ...Array(chunkSize / 2).fill(' '));\n // Pad end to make length a multiple of chunkSize\n const padding2Length = (chunkSize - docChars2.length % chunkSize) % chunkSize;\n docChars2.push(...Array(padding2Length).fill(' '));\n const fragments2 = docChars2.reduce(getFragmenter(chunkSize, {symbolsArray: docChars, initialOffset: -(chunkSize/2)}), []); // Adjust offset\n\n DEBUG.verboseSlow && console.log(\"highlight: fragments1 count:\", fragments1.length, \"fragments2 count:\", fragments2.length);\n\n const allFragments = [...fragments1, ...fragments2];\n const scoredFragments = allFragments.map(fragment => {\n // fragment.text is already from the original doc, no need to lowercase it here for distance calculation\n // ukkonen should compare queryLower with fragment.text.toLocaleLowerCase()\n const distance = ukkonen(queryLower, fragment.text.toLocaleLowerCase(), maxDistance);\n \n let scaledScore;\n if (distance === -1) { // Exceeded maxDistance\n scaledScore = Infinity;\n } else {\n scaledScore = (distance - minPossibleScore) / maxScoreRange;\n }\n return {score: scaledScore, fragment}; // fragment object contains {text, offset, symbols}\n });\n\n // Sort ascending (smallest scores win)\n scoredFragments.sort((a, b) => a.score - b.score);\n\n const initialHighlights = [];\n for (const {score, fragment} of scoredFragments) {\n if (score > maxAcceptScore) {\n // If we already have some highlights, we can stop if scores get too bad.\n // If we have none, we might continue to find at least one, even if poor.\n if (initialHighlights.length > 0) break; \n }\n initialHighlights.push({score, fragment});\n if (initialHighlights.length >= 10 + 1) break; // Get a bit more than needed for the \"better\" selection (original took 10 for \"better\")\n }\n \n DEBUG.verboseSlow && console.log(\"highlight: initialHighlights count:\", initialHighlights.length);\n\n let topSnippets;\n\n if (initialHighlights.length === 0) {\n DEBUG.verboseSlow && console.log('highlight: Zero initial highlights. Considering first scored fragment if available.');\n // Original logic: scores.slice(0,1) - this implies taking the best raw score if no \"good\" highlights\n if (scoredFragments.length > 0 && scoredFragments[0].score !== Infinity) {\n // Take the single best fragment, expand context, and mark it.\n const bestFragment = scoredFragments[0].fragment;\n const contextChars = chunkSize; // Original 'extra' was chunkSize\n const start = Math.max(0, bestFragment.offset - contextChars);\n const end = Math.min(docChars.length, bestFragment.offset + Array.from(bestFragment.text).length + contextChars);\n const snippetText = docChars.slice(start, end).join('');\n \n topSnippets = [{\n // score: scoredFragments[0].score, // Keep score if needed\n fragment: {\n text: internalMarkText(snippetText, query),\n offset: bestFragment.offset // Original offset of the core matched chunk\n }\n }];\n } else {\n topSnippets = []; // Truly no usable fragments\n }\n } else {\n // --- \"Better\" loop for context expansion and re-scoring (Identical logic to original) ---\n const contextCharsForBetterLoop = chunkSize; // Original 'extra' was chunkSize\n let betterScoredSnippets = initialHighlights.slice(0, 10).map(hl => {\n const originalFragment = hl.fragment;\n const originalFragmentTextChars = Array.from(originalFragment.text); // Unicode safe length\n const originalFragmentLength = originalFragmentTextChars.length;\n\n // Expand context using original document characters (hl.fragment.symbols)\n const startContext = Math.max(0, originalFragment.offset - contextCharsForBetterLoop);\n const endContext = Math.min(originalFragment.symbols.length, originalFragment.offset + originalFragmentLength + contextCharsForBetterLoop);\n \n const expandedText = originalFragment.symbols.slice(startContext, endContext).join('');\n const expandedTextLength = Array.from(expandedText).length; // Unicode safe\n\n // Re-calculate Ukkonen parameters for this new expanded text against the query\n const {\n maxDistance: newMaxDist, \n minPossibleScore: newMinScore, \n maxScoreRange: newMaxScoreRange\n } = calculateUkkonenParams(queryLength, expandedTextLength); // chunkSize is now expandedTextLength\n\n const newDistance = ukkonen(queryLower, expandedText.toLocaleLowerCase(), newMaxDist);\n \n let newScaledScore;\n if (newDistance === -1) {\n newScaledScore = Infinity;\n } else {\n newScaledScore = (newDistance - newMinScore) / newMaxScoreRange;\n }\n \n // The fragment text for output is the expanded text\n return {\n score: newScaledScore, \n fragment: { // New fragment object\n text: expandedText, // This text will be marked later\n // The offset should ideally be the start of this expanded snippet in the original document\n offset: startContext, \n // symbols: originalFragment.symbols // Not needed in final output\n }\n };\n });\n\n betterScoredSnippets.sort((a, b) => a.score - b.score);\n DEBUG.verboseSlow && console.log(\"highlight: betterScoredSnippets (after re-scoring with context):\", JSON.stringify(betterScoredSnippets.slice(0,3),null,2));\n \n // Take top 3 from these \"better\" snippets and apply marking\n topSnippets = betterScoredSnippets.slice(0, 3).map(item => ({\n // score: item.score, // Keep score if needed\n fragment: {\n text: internalMarkText(item.fragment.text, query),\n offset: item.fragment.offset\n }\n }));\n }\n \n DEBUG.verboseSlow && console.log(\"highlight: final topSnippets to return:\", topSnippets);\n return topSnippets;\n}\n\n\n// --- getFragmenter (Helper for highlight and trilight) ---\n// Preserving its original logic as much as possible, with clearer parameters.\n// The `symbolsArray` and `initialOffset` are for `highlight`'s specific needs.\nfunction getFragmenter(chunkSize, {overlap = false, symbolsArray = null, initialOffset = 0} = {}) {\n if (!Number.isInteger(chunkSize) || chunkSize < 1) {\n throw new TypeError(`chunkSize needs to be a whole number greater than 0`);\n }\n\n let currentFragmentCharCount; // Renamed from currentLength for clarity\n\n return function fragmentReducer(fragmentsAccumulator, nextCharSymbol, charIndex, fullSymbolArray) {\n // `fullSymbolArray` is the array being reduced.\n // `symbolsArray` (passed in options) is the *original* document characters,\n // used by `highlight` to ensure fragment.symbols points to the original doc.\n const effectiveSymbolsArray = symbolsArray || fullSymbolArray;\n const effectiveCharIndex = charIndex + initialOffset; // Adjust index for highlight's second pass\n\n if (overlap) {\n // Logic for overlapping fragments (primarily for trilight's n-grams)\n // This part of original getFragmenter was complex and seemed to modify previous frags.\n // For n-grams, it's simpler: create a new fragment for each possible n-gram.\n if (charIndex <= fullSymbolArray.length - chunkSize) {\n const ngramChars = fullSymbolArray.slice(charIndex, charIndex + chunkSize);\n fragmentsAccumulator.push({\n text: ngramChars.join(''),\n offset: effectiveCharIndex, // Offset in the original document\n symbols: effectiveSymbolsArray\n });\n }\n } else {\n // Logic for non-overlapping fragments (for highlight's chunking)\n if (fragmentsAccumulator.length === 0 || currentFragmentCharCount >= chunkSize) {\n // Start a new fragment\n fragmentsAccumulator.push({\n text: nextCharSymbol,\n offset: effectiveCharIndex, // Offset in the original document\n symbols: effectiveSymbolsArray\n });\n currentFragmentCharCount = 1;\n } else {\n // Add to the current fragment\n const currentFragment = fragmentsAccumulator[fragmentsAccumulator.length - 1];\n currentFragment.text += nextCharSymbol;\n currentFragmentCharCount++;\n }\n }\n return fragmentsAccumulator;\n };\n}\n\n\n// --- trilight function ---\n// Preserving original algorithm and segment generation logic with clarity and support.\nexport function trilight(query, docString, {\n maxLength = 0,\n ngramSize = 3,\n maxSegmentSize = 140,\n // numResults is implicitly 3 due to .slice(0,3) at the end\n} = {}) {\n const originalDocChars = Array.from(docString); // For final slicing, Unicode safe\n const queryChars = Array.from(query.toLocaleLowerCase()); // Lowercase query once\n \n let docCharsForProcessing = Array.from(docString.toLocaleLowerCase());\n if (maxLength > 0 && docCharsForProcessing.length > maxLength) {\n docCharsForProcessing = docCharsForProcessing.slice(0, maxLength);\n }\n\n if (docCharsForProcessing.length < ngramSize || queryChars.length < ngramSize) {\n return [];\n }\n\n // Generate n-grams for document and query using the getFragmenter\n // For n-grams, getFragmenter should be called with overlap: true\n const docNgrams = docCharsForProcessing.reduce(getFragmenter(ngramSize, {overlap: true, symbolsArray: originalDocChars}), []);\n const queryNgrams = queryChars.reduce(getFragmenter(ngramSize, {overlap: true, symbolsArray: queryChars}), []); // symbolsArray here is queryChars\n\n if (docNgrams.length === 0 || queryNgrams.length === 0) return [];\n\n // Index document n-grams by their text\n const docNgramIndex = new Map();\n docNgrams.forEach(ngram => {\n if (!docNgramIndex.has(ngram.text)) {\n docNgramIndex.set(ngram.text, []);\n }\n // Store original character offset of the ngram in the document\n docNgramIndex.get(ngram.text).push(ngram.offset);\n });\n\n // --- Find matching entries (Identical to original logic) ---\n const matchingEntries = [];\n queryNgrams.forEach((queryNgram, queryNgramIndex) => {\n const docOffsetsForNgram = docNgramIndex.get(queryNgram.text);\n if (docOffsetsForNgram) {\n docOffsetsForNgram.forEach(docCharOffset => {\n matchingEntries.push({\n ngramText: queryNgram.text,\n queryNgramIndex: queryNgramIndex, // Index of ngram within queryNgrams list\n docCharOffset: docCharOffset // Character offset of ngram in original document\n });\n });\n }\n });\n matchingEntries.sort((a, b) => a.docCharOffset - b.docCharOffset); // Sort by document offset\n\n // --- Identify runs of consecutive matching n-grams (Identical to original logic) ---\n const runs = [];\n if (matchingEntries.length > 0) {\n let currentRun = {\n ngramsInRun: [matchingEntries[0].ngramText],\n startQueryNgramIndex: matchingEntries[0].queryNgramIndex,\n startDocCharOffset: matchingEntries[0].docCharOffset\n };\n let lastQueryNgramIndexInRun = matchingEntries[0].queryNgramIndex;\n let lastDocCharOffsetInRun = matchingEntries[0].docCharOffset;\n\n for (let i = 1; i < matchingEntries.length; i++) {\n const entry = matchingEntries[i];\n const queryIndexDiff = entry.queryNgramIndex - lastQueryNgramIndexInRun;\n const docOffsetDiff = entry.docCharOffset - lastDocCharOffsetInRun;\n\n if (queryIndexDiff === 1 && docOffsetDiff === 1) { // Consecutive in both query and doc\n currentRun.ngramsInRun.push(entry.ngramText);\n } else {\n // End current run, add its length, then push\n currentRun.charLengthInDoc = currentRun.ngramsInRun.length + (ngramSize - 1);\n runs.push(currentRun);\n // Start new run\n currentRun = {\n ngramsInRun: [entry.ngramText],\n startQueryNgramIndex: entry.queryNgramIndex,\n startDocCharOffset: entry.docCharOffset\n };\n }\n lastQueryNgramIndexInRun = entry.queryNgramIndex;\n lastDocCharOffsetInRun = entry.docCharOffset;\n }\n // Add the last run\n currentRun.charLengthInDoc = currentRun.ngramsInRun.length + (ngramSize - 1);\n runs.push(currentRun);\n }\n \n DEBUG.verboseSlow && console.log(\"trilight: identified runs:\", runs.length);\n\n // --- Calculate gaps between runs (Identical to original logic) ---\n const gaps = [];\n if (runs.length > 1) {\n for (let i = 0; i < runs.length - 1; i++) {\n const run1 = runs[i];\n const run2 = runs[i+1];\n gaps.push({\n connectedRuns: [run1, run2],\n gapSize: run2.startDocCharOffset - (run1.startDocCharOffset + run1.charLengthInDoc)\n });\n }\n }\n gaps.sort((a, b) => a.gapSize - b.gapSize); // Sort by smallest gap\n\n // --- Merge runs into segments (Identical to original logic) ---\n const segments = [];\n const runToSegmentMap = new Map(); // Maps run's startDocCharOffset to the segment it belongs to\n\n // Initialize segments with individual runs if they are not too long\n runs.forEach(run => {\n if (run.charLengthInDoc <= maxSegmentSize) {\n const newSegment = {\n startOffset: run.startDocCharOffset,\n endOffset: run.startDocCharOffset + run.charLengthInDoc,\n score: run.charLengthInDoc // Initial score is its own length\n };\n segments.push(newSegment);\n runToSegmentMap.set(run.startDocCharOffset, newSegment);\n }\n });\n\n\n for (const gapInfo of gaps) {\n const runLeft = gapInfo.connectedRuns[0];\n const runRight = gapInfo.connectedRuns[1];\n\n const segmentForLeftRun = runToSegmentMap.get(runLeft.startDocCharOffset);\n const segmentForRightRun = runToSegmentMap.get(runRight.startDocCharOffset);\n\n if (segmentForLeftRun && segmentForRightRun && segmentForLeftRun === segmentForRightRun) {\n continue; // Already in the same segment\n }\n\n let merged = false;\n if (segmentForLeftRun && !segmentForRightRun) { // Try to extend left segment with right run\n const potentialNewEnd = runRight.startDocCharOffset + runRight.charLengthInDoc;\n if ((potentialNewEnd - segmentForLeftRun.startOffset) <= maxSegmentSize) {\n segmentForLeftRun.endOffset = potentialNewEnd;\n segmentForLeftRun.score += runRight.charLengthInDoc; // Add length of right run\n runToSegmentMap.set(runRight.startDocCharOffset, segmentForLeftRun); // Right run now points to left's segment\n // Remove standalone segment for right run if it existed (it shouldn't if !segmentForRightRun)\n const rightRunStandaloneSegmentIndex = segments.findIndex(s => s.startOffset === runRight.startDocCharOffset && s.endOffset === runRight.startDocCharOffset + runRight.charLengthInDoc);\n if (rightRunStandaloneSegmentIndex > -1) segments.splice(rightRunStandaloneSegmentIndex, 1);\n merged = true;\n }\n } else if (!segmentForLeftRun && segmentForRightRun) { // Try to extend right segment with left run\n const potentialNewStart = runLeft.startDocCharOffset;\n if ((segmentForRightRun.endOffset - potentialNewStart) <= maxSegmentSize) {\n segmentForRightRun.startOffset = potentialNewStart;\n segmentForRightRun.score += runLeft.charLengthInDoc;\n runToSegmentMap.set(runLeft.startDocCharOffset, segmentForRightRun);\n const leftRunStandaloneSegmentIndex = segments.findIndex(s => s.startOffset === runLeft.startDocCharOffset && s.endOffset === runLeft.startDocCharOffset + runLeft.charLengthInDoc);\n if (leftRunStandaloneSegmentIndex > -1) segments.splice(leftRunStandaloneSegmentIndex, 1);\n merged = true;\n }\n } else if (segmentForLeftRun && segmentForRightRun) { // Both runs are in existing (different) segments, try to merge these segments\n const potentialNewLength = segmentForRightRun.endOffset - segmentForLeftRun.startOffset;\n if (potentialNewLength <= maxSegmentSize) {\n segmentForLeftRun.endOffset = segmentForRightRun.endOffset;\n segmentForLeftRun.score += segmentForRightRun.score; // Combine scores\n\n // All runs that were part of segmentForRightRun now point to segmentForLeftRun\n for (const [runStartOffset, seg] of runToSegmentMap.entries()) {\n if (seg === segmentForRightRun) {\n runToSegmentMap.set(runStartOffset, segmentForLeftRun);\n }\n }\n // Remove segmentForRightRun from segments array\n const rightSegmentIndex = segments.indexOf(segmentForRightRun);\n if (rightSegmentIndex > -1) segments.splice(rightSegmentIndex, 1);\n merged = true;\n }\n }\n // Original code also had a case for creating a new segment from two runs not yet in segments.\n // This is covered by the initialization of segments with individual runs, and then merging.\n // The provided logic for merging was:\n // else { /* if (!leftSeg && !rightSeg) */\n // const newSegment = { start: runs[0].di, end: runs[0].di + runs[0].length + nextGap.gap + runs[1].length, score: runs[0].length + runs[1].length };\n // if ( newSegment.end - newSegment.start <= maxSegmentSize ) { runSegMap[runs[0].di] = newSegment; runSegMap[runs[1].di] = newSegment; segments.push(newSegment); assigned = newSegment; }\n // }\n // This specific \"else\" is tricky to map directly if segments are pre-initialized.\n // The current merging logic tries to extend existing segments. If two runs are not in segments\n // and their combined length (including gap) is <= maxSegmentSize, they should form a new segment.\n // This is implicitly handled if they were small enough to be individual segments initially and then get merged.\n // The key is that `runToSegmentMap` correctly tracks which segment a run belongs to.\n\n if (merged) {\n DEBUG.verboseSlow && console.log('trilight: Merged gap, new segment length:', segmentForLeftRun ? segmentForLeftRun.endOffset - segmentForLeftRun.startOffset : segmentForRightRun.endOffset - segmentForRightRun.startOffset);\n } else {\n DEBUG.verboseSlow && console.log('trilight: Gap could not be merged or runs not in mappable segments.');\n }\n }\n \n // Deduplicate segments that might have become identical after merges (e.g., if map pointed multiple runs to same segment object)\n const uniqueSegments = Array.from(new Set(segments.filter(s => s))); // Filter out undefined/null if any\n uniqueSegments.sort((a, b) => b.score - a.score); // Sort by score (descending)\n\n const textSegments = uniqueSegments.slice(0, 3).map(segment => {\n const snippetText = originalDocChars.slice(segment.startOffset, segment.endOffset).join('');\n return { // Return in the same format as highlight()\n fragment: {\n text: internalMarkText(snippetText, query),\n offset: segment.startOffset\n }\n };\n });\n\n DEBUG.verboseSlow && console.log(\"trilight: final textSegments:\", textSegments.length);\n\n if (textSegments.length === 0 && originalDocChars.length > 0) {\n DEBUG.verboseSlow && console.log(\"trilight: No segments found, returning beginning of doc.\");\n const fallbackText = originalDocChars.slice(0, Math.min(maxSegmentSize, originalDocChars.length)).join('');\n return [{ fragment: { text: internalMarkText(fallbackText, query), offset: 0 } }];\n }\n\n return textSegments;\n}\n" }, { "path": "src/index.js", "content": " \nrequire = require('esm')(module/*, options*/);\nmodule.exports = require('./app.js');\n \n" }, { "path": "src/installBrowser.js", "content": "import { exec } from 'child_process';\nimport { promisify } from 'util';\nimport { createWriteStream } from 'fs';\nimport { pipeline } from 'stream/promises';\nimport { readFile } from 'fs/promises';\nimport https from 'node:https';\n\n// Constants\nconst execPromise = promisify(exec);\n\nconst SUPPORTED_BROWSERS = ['chrome', 'brave', 'vivaldi', 'edge', 'chromium'];\nconst PLATFORM = process.platform; // 'win32', 'darwin', 'linux'\nconst ARCH = process.arch; // 'x64', 'arm64', etc.\n\n// Logic\n// None; function is exported for use\n\n// Functions\nexport async function installBrowser(browserName) {\n if (!SUPPORTED_BROWSERS.includes(browserName)) {\n throw new Error(`Unsupported browser: ${browserName}. Supported: ${SUPPORTED_BROWSERS.join(', ')}`);\n }\n\n console.log(`Installing ${browserName} on ${PLATFORM} (${ARCH})...`);\n\n await checkBrowserAvailability(browserName);\n const binaryPath = await installBrowserForPlatform(browserName);\n console.log(`${browserName} installed at: ${binaryPath}`);\n return binaryPath;\n}\n\nasync function checkBrowserAvailability(browserName) {\n if (PLATFORM === 'linux' && ARCH === 'arm64' && browserName === 'chrome') {\n throw new Error('Chrome is not available for ARM64 Linux. Try Brave or Chromium instead.');\n }\n // Add more checks for other browsers if needed (e.g., Vivaldi ARM64 stability)\n}\n\nasync function installBrowserForPlatform(browserName) {\n if (PLATFORM === 'win32') {\n return await installOnWindows(browserName);\n } else if (PLATFORM === 'darwin') {\n return await installOnMacOS(browserName);\n } else if (PLATFORM === 'linux') {\n return await installOnLinux(browserName);\n } else {\n throw new Error(`Unsupported platform: ${PLATFORM}`);\n }\n}\n\nasync function installOnWindows(browserName) {\n try {\n // Check if winget is installed\n try {\n await execPromise('winget --version');\n } catch {\n console.log('winget not found. Installing winget...');\n await execPromise('powershell -Command \"irm asheroto.com/winget | iex\"');\n console.log('winget installed successfully.');\n }\n\n if (browserName === 'chrome') {\n await execPromise('winget install Google.Chrome --silent --accept-package-agreements --accept-source-agreements');\n return 'C:\\\\Program Files\\\\Google\\\\Chrome\\\\Application\\\\chrome.exe';\n } else if (browserName === 'brave') {\n await execPromise('winget install Brave.Brave --silent --accept-package-agreements --accept-source-agreements');\n return 'C:\\\\Program Files\\\\BraveSoftware\\\\Brave-Browser\\\\Application\\\\brave.exe';\n } else if (browserName === 'vivaldi') {\n await execPromise('winget install Vivaldi.Vivaldi --silent --accept-package-agreements --accept-source-agreements');\n return 'C:\\\\Program Files\\\\Vivaldi\\\\Application\\\\vivaldi.exe';\n } else if (browserName === 'edge') {\n await execPromise('winget install Microsoft.Edge --silent --accept-package-agreements --accept-source-agreements');\n return 'C:\\\\Program Files (x86)\\\\Microsoft\\\\Edge\\\\Application\\\\msedge.exe';\n } else if (browserName === 'chromium') {\n const url = getDownloadUrl(browserName, PLATFORM, ARCH);\n if (!url) throw new Error('Chromium download not supported on Windows');\n const outputPath = 'C:\\\\Program Files\\\\Chromium\\\\chromium.exe';\n await downloadBinary(url, outputPath);\n return outputPath;\n }\n } catch (error) {\n console.error(`Windows install failed: ${error.message}`);\n throw error;\n }\n}\n\nasync function installOnMacOS(browserName) {\n try {\n // Check if brew is installed\n try {\n await execPromise('brew --version');\n } catch {\n throw new Error('Homebrew is not installed. Please install it from https://brew.sh and try again.');\n }\n\n if (browserName === 'chrome') {\n await execPromise('brew install --cask google-chrome');\n return '/Applications/Google Chrome.app/Contents/MacOS/Google Chrome';\n } else if (browserName === 'brave') {\n await execPromise('brew install --cask brave-browser');\n return '/Applications/Brave Browser.app/Contents/MacOS/Brave Browser';\n } else if (browserName === 'vivaldi') {\n await execPromise('brew install --cask vivaldi');\n return '/Applications/Vivaldi.app/Contents/MacOS/Vivaldi';\n } else if (browserName === 'edge') {\n await execPromise('brew install --cask microsoft-edge');\n return '/Applications/Microsoft Edge.app/Contents/MacOS/Microsoft Edge';\n } else if (browserName === 'chromium') {\n await execPromise('brew install --cask chromium');\n return '/Applications/Chromium.app/Contents/MacOS/Chromium';\n }\n } catch (error) {\n console.error(`macOS install failed: ${error.message}`);\n throw error;\n }\n}\n\nasync function installOnLinux(browserName) {\n try {\n const distro = await getLinuxDistro();\n if (distro === 'debian') {\n return await installOnDebian(browserName);\n } else if (distro === 'fedora') {\n return await installOnFedora(browserName);\n } else {\n throw new Error(`Unsupported Linux distribution: ${distro}`);\n }\n } catch (error) {\n console.error(`Linux install failed: ${error.message}`);\n throw error;\n }\n}\n\nasync function installOnDebian(browserName) {\n let binaryPath = '/usr/bin/' + browserName;\n if (browserName === 'chrome') {\n await execPromise('wget -q -O - https://dl.google.com/linux/linux_signing_key.pub | sudo apt-key add -');\n await execPromise(`sudo sh -c 'echo \"deb [arch=${ARCH === 'arm64' ? 'arm64' : 'amd64'}] http://dl.google.com/linux/chrome/deb/ stable main\" >> /etc/apt/sources.list.d/google-chrome.list'`);\n await execPromise('sudo apt-get update && sudo apt-get install -y google-chrome-stable');\n binaryPath = '/usr/bin/google-chrome';\n } else if (browserName === 'brave') {\n await execPromise('sudo curl -fsSLo /usr/share/keyrings/brave-browser-archive-keyring.gpg https://brave-browser-apt-release.s3.brave.com/brave-browser-archive-keyring.gpg');\n await execPromise(`echo \"deb [signed-by=/usr/share/keyrings/brave-browser-archive-keyring.gpg arch=${ARCH === 'arm64' ? 'arm64' : 'amd64'}] https://brave-browser-apt-release.s3.brave.com/ stable main\" | sudo tee /etc/apt/sources.list.d/brave-browser-release.list`);\n await execPromise('sudo apt update && sudo apt install -y brave-browser');\n binaryPath = '/usr/bin/brave-browser';\n } else if (browserName === 'vivaldi') {\n await execPromise('wget -qO- https://repo.vivaldi.com/archive/linux_signing_key.pub | sudo apt-key add -');\n await execPromise(`sudo add-apt-repository \"deb [arch=${ARCH === 'arm64' ? 'arm64' : 'amd64'}] https://repo.vivaldi.com/archive/deb/ stable main\"`);\n await execPromise('sudo apt update && sudo apt install -y vivaldi-stable');\n binaryPath = '/usr/bin/vivaldi';\n } else if (browserName === 'edge') {\n await execPromise('curl https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor > microsoft.gpg');\n await execPromise('sudo mv microsoft.gpg /usr/share/keyrings/microsoft-archive-keyring.gpg');\n await execPromise(`sudo sh -c 'echo \"deb [arch=${ARCH === 'arm64' ? 'arm64' : 'amd64'} signed-by=/usr/share/keyrings/microsoft-archive-keyring.gpg] https://packages.microsoft.com/repos/edge stable main\" > /etc/apt/sources.list.d/microsoft-edge.list'`);\n await execPromise('sudo apt update && sudo apt install -y microsoft-edge-stable');\n binaryPath = '/usr/bin/microsoft-edge';\n } else if (browserName === 'chromium') {\n await execPromise('sudo apt update && sudo apt install -y chromium-browser');\n // Check for Snap installation\n try {\n const { stdout } = await execPromise('which chromium');\n binaryPath = stdout.trim();\n if (binaryPath.includes('/snap/')) {\n binaryPath = '/snap/bin/chromium';\n } else {\n binaryPath = '/usr/bin/chromium-browser';\n }\n } catch {\n binaryPath = '/usr/bin/chromium-browser';\n }\n }\n return binaryPath;\n}\n\nasync function installOnFedora(browserName) {\n let binaryPath = '/usr/bin/' + browserName;\n if (browserName === 'chrome') {\n await execPromise('sudo dnf config-manager --add-repo https://dl.google.com/linux/chrome/rpm/stable/x86_64');\n await execPromise('sudo rpm --import https://dl.google.com/linux/linux_signing_key.pub');\n await execPromise('sudo dnf install -y google-chrome-stable');\n binaryPath = '/usr/bin/google-chrome';\n } else if (browserName === 'brave') {\n await execPromise('sudo dnf config-manager --add-repo https://brave-browser-rpm-release.s3.brave.com/x86_64/');\n await execPromise('sudo rpm --import https://brave-browser-rpm-release.s3.brave.com/brave-core.asc');\n await execPromise('sudo dnf install -y brave-browser');\n binaryPath = '/usr/bin/brave-browser';\n } else if (browserName === 'vivaldi') {\n await execPromise('sudo dnf config-manager --add-repo https://repo.vivaldi.com/archive/vivaldi-fedora.repo');\n await execPromise('sudo dnf install -y vivaldi-stable');\n binaryPath = '/usr/bin/vivaldi';\n } else if (browserName === 'edge') {\n await execPromise('sudo rpm --import https://packages.microsoft.com/keys/microsoft.asc');\n await execPromise('sudo dnf config-manager --add-repo https://packages.microsoft.com/yumrepos/edge');\n await execPromise('sudo dnf install -y microsoft-edge-stable');\n binaryPath = '/usr/bin/microsoft-edge';\n } else if (browserName === 'chromium') {\n await execPromise('sudo dnf install -y chromium');\n binaryPath = '/usr/bin/chromium-browser';\n }\n return binaryPath;\n}\n\nasync function getLinuxDistro() {\n try {\n const osRelease = await readFile('/etc/os-release', 'utf8');\n const lines = osRelease.split('\\n');\n const releaseInfo = {};\n for (const line of lines) {\n const [key, value] = line.split('=');\n if (key && value) {\n releaseInfo[key] = value.replace(/\"/g, '');\n }\n }\n\n if (releaseInfo.ID === 'fedora' || releaseInfo.ID_LIKE?.includes('fedora')) {\n return 'fedora';\n } else if (releaseInfo.ID === 'debian' || releaseInfo.ID === 'ubuntu' || releaseInfo.ID_LIKE?.includes('debian')) {\n return 'debian';\n } else {\n return releaseInfo.ID || 'unknown';\n }\n } catch (error) {\n console.error(`Failed to read /etc/os-release: ${error.message}`);\n return 'unknown';\n }\n}\n\n// Helper functions\nasync function downloadBinary(url, outputPath) {\n const response = await fetch(url, {\n agent: url.startsWith('https:') ? new https.Agent({ keepAlive: true }) : undefined\n });\n if (!response.ok) {\n throw new Error(`Failed to download ${url}: ${response.statusText}`);\n }\n await pipeline(response.body, createWriteStream(outputPath));\n}\n\nfunction getDownloadUrl(browserName, platform, arch) {\n const urls = {\n chrome: {\n win32: { x64: 'https://dl.google.com/chrome/install/ChromeSetup.exe', arm64: 'https://dl.google.com/chrome/install/ChromeSetup.exe' },\n darwin: { x64: 'https://dl.google.com/chrome/mac/stable/GGRO/googlechrome.dmg', arm64: 'https://dl.google.com/chrome/mac/arm64/googlechrome.dmg' },\n linux: { x64: 'https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb', arm64: null }\n },\n brave: {\n win32: { x64: 'https://referrals.brave.com/latest/BraveBrowserSetup.exe', arm64: 'https://referrals.brave.com/latest/BraveBrowserSetup.exe' },\n darwin: { x64: 'https://laptop-updates.brave.com/latest/osx/Brave-Browser.dmg', arm64: 'https://laptop-updates.brave.com/latest/osx-arm64/Brave-Browser.dmg' },\n linux: { x64: 'https://laptop-updates.brave.com/latest/linux64', arm64: 'https://laptop-updates.brave.com/latest/linux-arm64' }\n },\n vivaldi: {\n win32: { x64: 'https://downloads.vivaldi.com/stable/Vivaldi_Setup.exe', arm64: 'https://downloads.vivaldi.com/stable/Vivaldi_Setup.exe' },\n darwin: { x64: 'https://downloads.vivaldi.com/stable/Vivaldi.dmg', arm64: 'https://downloads.vivaldi.com/stable/Vivaldi.dmg' },\n linux: { x64: 'https://downloads.vivaldi.com/stable/vivaldi-stable_amd64.deb', arm64: 'https://downloads.vivaldi.com/stable/vivaldi-stable_arm64.deb' }\n },\n edge: {\n win32: { x64: 'https://go.microsoft.com/fwlink/?linkid=2069148', arm64: 'https://go.microsoft.com/fwlink/?linkid=2069148' },\n darwin: { x64: 'https://go.microsoft.com/fwlink/?linkid=2069324', arm64: 'https://go.microsoft.com/fwlink/?linkid=2069324' },\n linux: { x64: 'https://packages.microsoft.com/repos/edge/pool/main/m/microsoft-edge-stable/microsoft-edge-stable_latest_amd64.deb', arm64: 'https://packages.microsoft.com/repos/edge/pool/main/m/microsoft-edge-stable/microsoft-edge-stable_latest_arm64.deb' }\n },\n chromium: {\n win32: { x64: 'https://download-chromium.appspot.com/dl/Win_x64?type=snapshots', arm64: 'https://download-chromium.appspot.com/dl/Win_arm64?type=snapshots' },\n darwin: { x64: 'https://download-chromium.appspot.com/dl/Mac?type=snapshots', arm64: 'https://download-chromium.appspot.com/dl/Mac_Arm?type=snapshots' },\n linux: { x64: 'https://download-chromium.appspot.com/dl/Linux_x64?type=snapshots', arm64: 'https://download-chromium.appspot.com/dl/Linux_Arm?type=snapshots' }\n }\n };\n return urls[browserName]?.[platform]?.[arch] || null;\n}\n" }, { "path": "src/launcher.js", "content": "// launcher.js\nimport { spawn } from 'child_process';\nimport { DEBUG } from './common.js'; // Assuming common.js is accessible\n\n/**\n * Launches a browser executable with specified arguments.\n * @param {string} executablePath - Absolute path to the browser executable.\n * @param {string[]} browserArgs - Array of arguments to pass to the browser.\n * @param {object} [options={}] - Options for child_process.spawn.\n * @returns {import('child_process').ChildProcess | null} The spawned browser process or null on error.\n */\nfunction launch(executablePath, browserArgs = [], options = {}) {\n if (!executablePath) {\n console.error('launcher.js: Executable path is required.');\n return null;\n }\n\n DEBUG.verbose && console.log(`launcher.js: Spawning '${executablePath}' with args:`, browserArgs);\n\n try {\n const defaultSpawnOptions = {\n detached: process.platform !== 'win32', // Detach by default on non-Windows for independent exit\n stdio: ['ignore', 'pipe', 'pipe'], \n };\n\n const spawnOptions = { ...defaultSpawnOptions, ...options };\n\n const browserProcess = spawn(executablePath, browserArgs, spawnOptions);\n\n browserProcess.on('error', (err) => {\n console.error(`launcher.js: Failed to start browser process for ${executablePath}: ${err.message}`);\n });\n\n if (DEBUG.verboseBrowser) {\n const browserName = executablePath.split(/[/\\\\]/).pop();\n browserProcess.stdout.on('data', (data) => {\n DEBUG.verbose && process.stdout.write(`[BROWSER STDOUT - ${browserName}]: ${data}`);\n });\n browserProcess.stderr.on('data', (data) => {\n DEBUG.verbose && process.stderr.write(`[BROWSER STDERR - ${browserName}]: ${data}`);\n });\n }\n \n // If detached, unref() allows the parent to exit independently.\n // This is often desired so closing the terminal doesn't kill the browser launched by the script.\n if (spawnOptions.detached) {\n browserProcess.unref();\n }\n\n return browserProcess;\n } catch (error) {\n console.error(`launcher.js: Error spawning browser ${executablePath}: ${error.message}`);\n DEBUG.verbose && console.error(error);\n return null;\n }\n}\n\nexport default {\n launch,\n};\n" }, { "path": "src/libraryServer.js", "content": "import sea from 'node:sea';\nimport http from 'http';\nimport https from 'https';\nimport fs from 'fs';\nimport os from 'os'; // Included as it was in your original list, though not directly used in this server logic\nimport path from 'path';\n\nimport express from 'express';\n\nimport args from './args.js';\nimport {\n GO_SECURE,\n CERT_PATH,\n DEBUG,\n MAX_REAL_URL_LENGTH,\n MAX_HEAD, MAX_HIGHLIGHTABLE_LENGTH,\n say, sleep, APP_ROOT,\n RichError\n} from './common.js';\nimport {startCrawl, Archivist} from './archivist.js';\nimport {trilight, highlight} from './highlighter.js'; // trilight is imported but its usage was commented out in original\n\nconst SITE_PATH = path.resolve(APP_ROOT, '..', 'public'); // Serves static files like style.css from here\n\nconst SearchCache = new Map();\n\nconst app = express();\n\nlet running = false;\nlet Server, upAt, port;\n\nconst LibraryServer = {\n start, stop\n}\n\nconst secure_options = {};\nconst protocol = GO_SECURE ? https : http;\n\nexport default LibraryServer;\n\n// --- PageLayout Helper ---\n// (Incorporates changes for new default page and /settings path)\nfunction PageLayout({ title, content, currentNav, layoutType = 'default' }) {\n return `\n \n \n \n \n \nYour index is currently empty. Start a crawl to add items!
' : ''}\nNo results found for your query.
' : ''}\n ${!query && results.length === 0 ? 'Please enter a search term above to begin.
' : ''}\n\n\n ${results.length > 0 ? `\nShowing results for: ${query}
\n${snippet}
` : ''}\n