ابحث بالوصف 🎬

٢٠٥٦٣ عملاً · ترشيح لفظي ثم إعادة ترتيب تفهم العربي · صفر مساس بحصصك

\n\n\n

\n\n

\n`;\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n if (u.pathname === \"/api\") {\n const q = (u.searchParams.get(\"q\") || \"\").trim();\n if (!q) return Response.json({ items: [], debug: {} });\n try { return Response.json(await api(q)); }\n catch (e) { return Response.json({ error: String(e).slice(0, 300) }); }\n }\n if (u.pathname === \"/chips\") {\n const k = (u.searchParams.get(\"k\") || \"\").split(\",\").map(s => s.trim()).filter(Boolean);\n if (k.length < 2) return Response.json({ error: \"اختر رقيقتين على الأقل\" });\n try { return Response.json(await byChips(k)); }\n catch (e) { return Response.json({ error: String(e).slice(0, 300) }); }\n }\n return new Response(HTML, { headers: { \"content-type\": \"text/html; charset=utf-8\" } });\n}\n",[],[422],"362987b1-d393-48c6-a3a4-29176042d89b","peoplebench.ts","ad4bab30-b131-11f1-b93f-1607ee4eb77e",188,["D",1789496234397],"/* قياس أثر عمود الطاقم: هل يصلح بحث الممثلين بلا إضرار بالباقي؟ */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\n/* حالات تخص الطاقم — الخطأ الذي نصلحه */\nconst ACTOR: [string, string][] = [\n [\"a movie with Brad Pitt about an underground fight club\", \"Fight Club\"],\n [\"a movie directed by Christopher Nolan about dreams\", \"Inception\"],\n [\"a movie with Tom Hanks about a man stranded on an island\", \"Cast Away\"],\n [\"a movie with Leonardo DiCaprio about a sinking ship\", \"Titanic\"],\n];\n\nconst W: [string, string][] = [\n [\"بلا طاقم\", \"bm25(plotP,1,1,1,0)\"],\n [\"طاقم ٠٫٥\", \"bm25(plotP,1,1,1,0.5)\"],\n [\"طاقم ١\", \"bm25(plotP,1,1,1,1)\"],\n [\"طاقم ٢\", \"bm25(plotP,1,1,1,2)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const mode = u.searchParams.get(\"mode\") || \"actor\";\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const scope = await sqlite.execute(`SELECT title FROM plotP`);\n const allowed = new Set((scope.rows || []).map((x: any) => String(x[0])));\n\n const list: [string, string][] = mode === \"actor\" ? ACTOR\n : CASES.filter(c => [...allowed].some(t => { const x = norm(t), w = norm(c[1]); return x === w || x.startsWith(w); })) as any;\n\n const tally: any = {}; for (const [n] of W) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n let tested = 0;\n\n for (const [ar, exp] of list.slice(0, 22)) {\n const want = norm(exp);\n if (![...allowed].some(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); })) continue;\n tested++;\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const parts: string[] = [];\n\n for (const [name, rank] of W) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM plotP WHERE plotP MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n const l = (r.rows || []).map((x: any) => String(x[0]));\n const rk = l.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n return Response.json({\n الوضع: mode === \"actor\" ? \"حالات الطاقم\" : \"الحالات العادية\",\n \"نطاق المقارنة\": allowed.size, مختبَر: tested,\n أعمدة: W.map(w => w[0]).join(\" · \"),\n النتيجة: Object.fromEntries(W.map(([n]) =>\n [n, `أول ${tally[n][0]}/${tested} · ضمن٥ ${tally[n][1]}/${tested} · ضمن٢٠ ${tally[n][2]}/${tested}`])),\n التفصيل: rows,\n });\n}\n",[],[430],"ee2859e3-19d3-4bc6-858c-ca5c88fb46a3","people.ts","2fb539a2-b131-11f1-a025-1607ee4eb77e",186,["D",1789496023700],"/* إثراء بالطاقم: الممثلون والمخرجون — نص قصير عالي التمييز (نفس نمط الوسوم الناجح) */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\n\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; await fn(items[k]); }\n }));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1500\");\n\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = all.slice(start, start + count).filter(r => r.imdb && !r.people);\n\n const groups: any[][] = [];\n for (let i = 0; i < slice.length; i += 5) groups.push(slice.slice(i, i + 5));\n\n const t0 = Date.now();\n let got = 0;\n await pool(groups, 8, async (g: any[]) => {\n const alias = g.map((r, i) => `t${i}: title(id: ${JSON.stringify(r.imdb)}) {\n principalCredits(filter: { categories: [\"cast\",\"director\",\"writer\"] }) {\n credits(limit: 8) { name { nameText { text } } } } }`).join(\" \");\n const res = await gql(`query { ${alias} }`);\n g.forEach((r, i) => {\n const pcs = (res?.data?.[`t${i}`]?.principalCredits) || [];\n const names: string[] = [];\n for (const pc of pcs) for (const c of (pc?.credits || [])) {\n const n = c?.name?.nameText?.text;\n if (n && !names.includes(n)) names.push(n);\n }\n if (names.length) { r.people = names.slice(0, 18).join(\", \"); got++; }\n });\n });\n\n await blob.setJSON(\"corpus_v2\", all);\n const done = all.filter(r => r.people).length;\n return Response.json({\n عولج: slice.length, نجح: got, ms: Date.now() - t0,\n \"لديه طاقم\": done, الإجمالي: all.length, التالي: start + count,\n عينة: (all.find(r => r.people)?.people || \"\").slice(0, 120),\n });\n}\n",[],[438],"046fffc1-6318-447d-b78c-3a21421b00a7","ondevice.ts","fdaf6144-affa-11f1-b05b-1607ee4eb77e",73,["D",1789362795764],"const CORPUS_URL = \"https://q8kamikaze--dfc53fe6affa11f198ef1607ee4eb77e.web.val.run/\";\n\nconst HTML = `\n\n\nقياس البحث على الجهاز\n\n\n

قياس البحث الدلالي على جهازك 🧪

النموذج يشتغل داخل جوالك — صفر خادم، صفر حساب، صفر حصة.\nالصفحة تقيس نفسها: زمن التحميل، زمن التجهيز، زمن البحث، ودقة ١٥ اختباراً حقيقياً.

\n\n

\n\n`;\n\nexport default () => new Response(HTML, { headers: { \"content-type\": \"text/html; charset=utf-8\" } });\n",[],[446],"2cd2bb01-27e9-4b8d-b240-112b032c9e13","mixbench.ts","e6c87546-b0c3-11f1-990e-1607ee4eb77e",150,["D",1789449086209],"/* القياس النهائي: هل الخلطة (حبكة ويكيبيديا + حلقات) ترفع الدقة فعلاً؟\n plot5 = عمودان (ملخّص+وسوم) · plot8 = ثلاثة (+حبكة) */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst W: [string, string, string][] = [\n [\"عمودان (الحالي)\", \"plot5\", \"bm25(plot5,1,1,1)\"],\n [\"+حبكة ١:١:١\", \"plot8\", \"bm25(plot8,1,1,1,1)\"],\n [\"+حبكة وزن ٢\", \"plot8\", \"bm25(plot8,1,1,1,2)\"],\n [\"+حبكة وزن ٣\", \"plot8\", \"bm25(plot8,1,1,1,3)\"],\n [\"حبكة فقط\", \"plot8\", \"bm25(plot8,1,0,0,1)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"15\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const tally: any = {}; for (const [n] of W) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const want = norm(exp);\n const parts: string[] = [];\n\n for (const [name, table, rank] of W) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY ${rank} LIMIT 20`,\n args: [match],\n });\n const list = (r.rows || []).map((x: any) => String(x[0]));\n const rk = list.findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: W.map(w => w[0]).join(\" · \"),\n النتيجة: Object.fromEntries(W.map(([name]) =>\n [name, `أول ${tally[name][0]}/${n} · ضمن٥ ${tally[name][1]}/${n} · ضمن٢٠ ${tally[name][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[454],"d9f04b43-f127-4594-854f-7233d1cbbc93","main.ts",55,["D",1789325062701],5,"import { findByPlot } from \"./engine.ts\";\nimport { findByKeywords } from \"./kwengine.ts\";\n\nconst HTML = `\n\n\n\n\nابحث بالوصف — تجربة حية\n\n\n\n\n

ابحث بالوصف — تجربة حية 🧪

\n \n \n

\n\n\n

مجاني ١٠٠٪: ترجمة ← SearXNG (سيرفرك) ← TMDB. ⚠️ يعتمد على محركات بحث تُعلَّق عند كثرة الطلبات.

\n \n \n

\n\n\n

الفكرة: وسوم كتبها بشر لكل عمل بـIMDb. نحوّل وصفك لوسوم، نتقاطعها، ثم نتحقّق من وسوم كل مرشَّح نفسه ونرتّب بعدد المطابق. صفر ذكاء اصطناعي · صفر محرك بحث · صفر حجب.

\n \n \n

\n \n

\n\n

\n\n\n`;\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n if (url.pathname === \"/api\") {\n const q = (url.searchParams.get(\"q\") || \"\").trim();\n if (!q) return Response.json({ error: \"empty\" });\n try { return Response.json(await findByPlot(q)); }\n catch (e) { return Response.json({ error: String(e).slice(0, 400) }); }\n }\n if (url.pathname === \"/kw\") {\n const q = (url.searchParams.get(\"q\") || \"\").trim();\n const pick = (url.searchParams.get(\"pick\") || \"\").split(\",\").map(s => s.trim()).filter(Boolean);\n if (!q) return Response.json({ error: \"empty\" });\n try { return Response.json(await findByKeywords(q, pick.length ? pick : undefined)); }\n catch (e) { return Response.json({ error: String(e).slice(0, 400) }); }\n }\n return new Response(HTML, { headers: { \"content-type\": \"text/html; charset=utf-8\" } });\n}\n",[],[462],"afe34866-1484-4fcc-ab0a-e35cf402481d","kwengine.ts","76f16f0c-afa1-11f1-8955-1607ee4eb77e",54,["D",1789324990359],14,"/* دراسة عمق — أنبوب وسوم IMDb البشرية + تصويت الأزواج\n ⚠️ ثلاثة أخطاء أمسكها القياس وصُحّحت هنا:\n ١) سقف الاتساع ٨٠٠٠ كان **يرمي أفضل وسم بالحالتين**: time-travel (٩١٥٥) وmonster (٩٥٩٥).\n السقف الآن ٣٠٬٠٠٠ — يرمي العام حقاً لا المميّز.\n ٢) المطابقة التقريبية (startsWith) أدخلت وسوماً دخيلة تماماً (\"theremin\") — أُلغيت، حرفي فقط.\n ٣) الترتيب بالندرة وحدها رفع مطابقات عرضية؛ الآن: العبارة الأطول أولاً (الأدق دلالياً)،\n ثم الأندر. */\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst SHASHTI = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst H = {\n \"content-type\": \"application/json\",\n \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\",\n};\nasync function gql(query: string) {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query }), signal: AbortSignal.timeout(25000) });\n return await r.json();\n}\n\n/* كلمات لا تصلح وسماً **مستقلاً** — لكنها مسموحة **داخل** وسم مركّب\n (\"time\" وحدها ضجيج، لكن \"time travel\" أقوى وسم ممكن. خلطهما كان خطأً مقاساً.) */\nconst WEAK = new Set((\"lives live living life time man woman boy girl child children thing place world year day night \" +\n \"hunt sound family people story\").split(\" \"));\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been about after \" +\n \"who that this his her its their he she it they them movie film series show tv season episode one two three \" +\n \"new old make made get gets has have had not no all some any more most where when how why \" +\n \"someone something himself herself themselves each other into onto over under while during because\").split(\" \"));\n\nconst slug = (s: string) => s.trim().toLowerCase().replace(/\\s+/g, \"-\");\n\nasync function toKeywords(text: string) {\n const words = text.toLowerCase().replace(/[^a-z0-9 ]+/g, \" \").split(/\\s+/).filter(Boolean);\n /* جذر خفيف: أول ٥ حروف — يجسر \"traveling\"↔\"travel\" و\"monsters\"↔\"monster\" */\n const stem = (w: string) => w.slice(0, 5);\n const bag = new Set(words.filter(w => !STOP.has(w) && w.length > 2).map(stem));\n\n const cand: string[] = [];\n for (let n = 3; n >= 1; n--) {\n for (let i = 0; i + n <= words.length; i++) {\n const g = words.slice(i, i + n);\n if (STOP.has(g[0]) || STOP.has(g[g.length - 1])) continue;\n if (g.every(w => STOP.has(w) || w.length <= 2)) continue;\n cand.push(g.join(\" \"));\n }\n }\n const uniq = [...new Set(cand)].slice(0, 24);\n /* مفرد/جمع: \"monsters\" ليست وسماً و\"monster\" وسم — نضيف المفرد كمرشَّح مستقل */\n const singular = uniq.map(w => w.replace(/\\b(\\w{4,})s\\b/g, \"$1\")).filter(w => !uniq.includes(w));\n const all = [...uniq, ...new Set(singular)].slice(0, 32);\n const hits = await Promise.all(all.map(async w => {\n try {\n const j = await gql(`query { mainSearch(first: 6, options: { searchTerm: ${JSON.stringify(w)}, type: KEYWORD }) {\n edges { node { entity { ... on Keyword { text { text } } } } } } }`);\n const list: string[] = (j?.data?.mainSearch?.edges || []).map((e: any) => e.node?.entity?.text?.text).filter(Boolean);\n const exact = list.find(h => h === w);\n /* ⚠️ تصحيح رابع (قياس): الترجمة تقلب ترتيب العبارة (\"traveling in time\" مقابل\n \"time travel\")، فالثنائية الحرفية لا تتكوّن أبداً ويضيع أقوى وسم. الحل:\n نقبل أي اقتراح **كل كلماته موجودة بالوصف نفسه** — لا شرط على الترتيب. */\n const subset = list\n .filter(h => h !== w && h.split(\" \").length > 1)\n .filter(h => h.split(\" \").every(x => STOP.has(x) || bag.has(stem(x))));\n const chosen = exact || subset[0] || null;\n return { w, kw: chosen, n: chosen ? chosen.split(\" \").length : 0, via: exact ? \"حرفي\" : (chosen ? \"تجميع\" : \"\") };\n } catch { return { w, kw: null, n: 0, via: \"\" }; }\n }));\n const seen = new Set();\n const strong: any[] = [], weak: any[] = [];\n for (const h of hits.filter(x => x.kw).sort((a, b) => b.n - a.n)) {\n const s = slug(h.kw!);\n if (seen.has(s)) continue;\n seen.add(s);\n /* كلمة ضعيفة مفردة تُؤجَّل احتياطاً — تُستعمل فقط لو الوسوم القوية أقل من ثلاثة */\n (h.n === 1 && WEAK.has(h.kw!.toLowerCase()) ? weak : strong).push({ ...h, slug: s });\n }\n return strong.length >= 3 ? strong : strong.concat(weak);\n}\n\nasync function keywordBreadth(s: string) {\n try {\n const j = await gql(`query { advancedTitleSearch(first: 1,\n constraints: { keywordConstraint: { allKeywords: ${JSON.stringify([s])} },\n explicitContentConstraint: { explicitContentFilter: EXCLUDE_ADULT } }) { total } }`);\n return j?.data?.advancedTitleSearch?.total ?? -1;\n } catch { return -1; }\n}\n\nasync function pair(a: string, b: string, take = 12) {\n try {\n const j = await gql(`query { advancedTitleSearch(first: ${take},\n sort: { sortBy: POPULARITY, sortOrder: ASC },\n constraints: { keywordConstraint: { allKeywords: ${JSON.stringify([a, b])} },\n explicitContentConstraint: { explicitContentFilter: EXCLUDE_ADULT } })\n { total edges { node { title { id titleText { text } releaseYear { year } titleType { id }\n primaryImage { url } ratingsSummary { aggregateRating } } } } } }`);\n const d = j?.data?.advancedTitleSearch;\n return { total: d?.total ?? 0, items: (d?.edges || []).map((e: any) => ({\n id: e.node?.title?.id, t: e.node?.title?.titleText?.text, y: e.node?.title?.releaseYear?.year,\n type: e.node?.title?.titleType?.id, img: e.node?.title?.primaryImage?.url || null,\n r: e.node?.title?.ratingsSummary?.aggregateRating ?? null,\n })) };\n } catch { return { total: 0, items: [] }; }\n}\n\nexport async function findByKeywords(rawText: string, picked?: string[]) {\n const t0 = Date.now();\n const dbg: any = {};\n let en = rawText.trim();\n if (/[\\u0600-\\u06FF]/.test(en)) {\n try {\n const tr = await fetch(SHASHTI + \"/translate?q=\" + encodeURIComponent(en) + \"&target=en\", { signal: AbortSignal.timeout(15000) })\n .then(r => r.ok ? r.json() : null);\n if (tr?.text) en = String(tr.text);\n } catch { /* نكمل بالأصل */ }\n }\n dbg.en = en;\n\n const kws = await toKeywords(en);\n dbg.mapped = kws.map(k => k.w + \" → \" + k.kw);\n\n const withBreadth = await Promise.all(kws.slice(0, 12).map(async k => ({ ...k, total: await keywordBreadth(k.slug) })));\n const auto = withBreadth\n .filter(k => k.total >= 5 && k.total <= 30000)\n .sort((a, b) => (b.n - a.n) || (a.total - b.total))\n .slice(0, 5);\n\n /* رقائق للعرض: كل الوسوم الصالحة + اقتراحات IMDb حولها (يجسر الفجوة اللفظية:\n المستخدم يكتب \"dream\" وIMDb يسمّيها \"dream within a dream\") */\n const sugSeeds = withBreadth.filter(k => k.total >= 5).slice(0, 4);\n const sugLists = await Promise.all(sugSeeds.map(async k => {\n try {\n const j = await gql(`query { mainSearch(first: 8, options: { searchTerm: ${JSON.stringify(k.kw)}, type: KEYWORD }) {\n edges { node { entity { ... on Keyword { text { text } } } } } } }`);\n return (j?.data?.mainSearch?.edges || []).map((e: any) => e.node?.entity?.text?.text).filter(Boolean);\n } catch { return []; }\n }));\n const chipSet = new Map();\n for (const k of withBreadth.filter(x => x.total >= 5)) chipSet.set(k.slug, true);\n for (const l of sugLists) for (const s of l) chipSet.set(slug(s), false);\n dbg.chips = [...chipSet.entries()].slice(0, 24).map(([s, isAuto]) => ({ s, auto: isAuto }));\n\n /* لو المستخدم اختار رقائق بنفسه، هي المصدر — وإلا الاختيار الآلي */\n const good = picked && picked.length >= 2\n ? picked.map(p => ({ slug: p, total: 0, n: p.split(\"-\").length, w: p, kw: p }))\n : auto;\n dbg.breadth = withBreadth.map(k => k.slug + \":\" + k.total);\n dbg.chosen = good.map((k: any) => k.slug + (k.total ? \"(\" + k.total + \")\" : \"\"));\n dbg.manual = !!(picked && picked.length >= 2);\n\n if (dbg.manual) {\n for (const g of good as any[]) if (!g.total) g.total = await keywordBreadth(g.slug);\n }\n\n if (good.length < 2) { dbg.ms = Date.now() - t0; return { items: [], debug: dbg, reason: \"وسوم غير كافية\" }; }\n\n const pairs: [string, string][] = [];\n for (let i = 0; i < good.length; i++)\n for (let j = i + 1; j < good.length; j++) pairs.push([good[i].slug, good[j].slug]);\n\n const results = await Promise.all(pairs.slice(0, 10).map(async ([a, b]) => ({ a, b, ...(await pair(a, b)) })));\n dbg.pairs = results.map(r => r.a + \"+\" + r.b + \"=\" + r.total);\n\n const votes = new Map();\n for (const r of results) {\n if (!r.total || r.total > 2000) continue;\n r.items.forEach((it: any) => {\n if (!it.id) return;\n let v = votes.get(it.id);\n if (!v) { v = { ...it, pairs: 0, from: [] as string[] }; votes.set(it.id, v); }\n v.pairs++;\n if (v.from.length < 4) v.from.push(r.a + \"+\" + r.b);\n });\n }\n\n /* ═══ الحل الجذري لفشل الاختيار الآلي ═══\n لا نحاول تخمين «أي وسمين هما الجوهريان» — نجمع كل المرشَّحين من كل الأزواج،\n ثم نسحب **قائمة وسوم كل مرشَّح نفسه** ونحسب كم وسماً من وسوم وصفك يحملها فعلاً.\n العمل الصحيح يحمل أربعة أو خمسة منها؛ الدخيل يحمل الاثنين اللذين جاء بهما فقط.\n والوزن IDF: وسم نادر (missing-children ٥٨) يساوي أضعاف وسم عام (time-travel ٩١٥٥). */\n const cands = [...votes.values()].slice(0, 24);\n if (cands.length) {\n const alias = cands.map((c, i) =>\n `t${i}: title(id: ${JSON.stringify(c.id)}) { keywords(first: 40) { edges { node { text } } } }`).join(\" \");\n try {\n const j = await gql(`query { ${alias} }`);\n const qk = withBreadth.filter(k => k.total > 0)\n .concat(dbg.manual ? (good as any[]).filter(g => !withBreadth.some(w => w.slug === g.slug)) : []);\n const BASE = 400000;\n const stemSet = (s: string) => new Set(s.split(\"-\").filter(x => x.length > 2 && !STOP.has(x)).map(x => x.slice(0, 5)));\n cands.forEach((c, i) => {\n const own = ((j?.data?.[`t${i}`]?.keywords?.edges) || []).map((e: any) => slug(String(e.node?.text || \"\")));\n const ownStems = own.map((o: string) => ({ o, st: stemSet(o) }));\n const used = new Set();\n let hit = 0, idf = 0;\n const matched: string[] = [];\n for (const k of qk) {\n const mine = stemSet(k.slug);\n if (!mine.size) continue;\n /* ⚠️ المطابقة الحرفية للوسم كانت تُسقط الصحيح: Dark يحمل \"missing child\"\n و\"german town\" بينما وصفك أنتج \"missing children\" و\"german\" — نفس المعنى،\n سلسلة مختلفة. فالمطابقة الآن بتداخل الجذور (أحدهما يحتوي الآخر). */\n let best: any = null, exact = false;\n for (const o of ownStems) {\n if (used.has(o.o)) continue;\n if (o.o === k.slug) { best = o; exact = true; break; }\n /* ⚠️ بگ مقاس: وسم بلا جذور معتبرة (k9 · ak-47) مجموعته فارغة،\n و«every» على مجموعة فارغة ترجع true — فينتج تطابق كاذب تماماً.\n الشرط الآن: الطرفان غير فارغين، ومع تداخل فعلي. */\n if (!o.st.size || !mine.size) continue;\n const sub = [...mine].every(x => o.st.has(x)) || [...o.st].every(x => mine.has(x));\n if (sub && !best) best = o;\n }\n if (!best) continue;\n used.add(best.o);\n hit++;\n idf += Math.log(BASE / Math.max(k.total, 5)) * (exact ? 1 : 0.7);\n matched.push(k.slug + (exact ? \"\" : \"≈\" + best.o));\n }\n c.hit = hit; c.idf = +idf.toFixed(2); c.matched = matched;\n });\n } catch (e) { dbg.verifyErr = String(e).slice(0, 120); }\n }\n\n const ranked = cands\n .sort((a, b) => (b.hit || 0) - (a.hit || 0) || (b.idf || 0) - (a.idf || 0))\n .slice(0, 8);\n dbg.ms = Date.now() - t0;\n return { items: ranked, debug: dbg };\n}\n",[],[470],"c47f5e33-3d3a-47de-bcee-4bd1fa2e8649","kwbench2.ts","66cdcccc-b0c5-11f1-a55b-1607ee4eb77e",154,["D",1789449730490],"/* قياس أثر توسيع الوسوم — مقارنة عادلة على نفس نطاق الألف عمل */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"45\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n /* نطاق المقارنة: الأعمال الموجودة بـplot9 فقط */\n const scope = await sqlite.execute(`SELECT title FROM plot9`);\n const allowed = new Set((scope.rows || []).map((x: any) => String(x[0])));\n\n const V = [\"وسوم ٢٥ (الحالي)\", \"وسوم ٨١ (موسّع)\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n let tested = 0;\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const want = norm(exp);\n if (![...allowed].some(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); })) continue;\n tested++;\n\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n\n const a = await sqlite.execute({\n sql: `SELECT title FROM plot5 WHERE plot5 MATCH ? ORDER BY bm25(plot5,1,1,1) LIMIT 300`, args: [match] });\n const b = await sqlite.execute({\n sql: `SELECT title FROM plot9 WHERE plot9 MATCH ? ORDER BY bm25(plot9,1,1,1) LIMIT 20`, args: [match] });\n\n const lists: Record = {\n \"وسوم ٢٥ (الحالي)\": (a.rows || []).map((x: any) => String(x[0])).filter(t => allowed.has(t)).slice(0, 20),\n \"وسوم ٨١ (موسّع)\": (b.rows || []).map((x: any) => String(x[0])),\n };\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" → \")}`);\n }\n\n return Response.json({\n \"نطاق المقارنة\": allowed.size + \" عملاً\",\n \"حالات مختبَرة\": tested,\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${tested} · ضمن٥ ${tally[v][1]}/${tested} · ضمن٢٠ ${tally[v][2]}/${tested}`])),\n التفصيل: rows,\n });\n}\n",[],[479],"0399d86d-3497-4b1c-bd7c-45a126c8f86b","kwbench.ts","7dc0073a-afa1-11f1-b4e3-1607ee4eb77e",47,["D",1789324674682],"import { findByKeywords } from \"./kwengine.ts\";\n\nconst CASES: [string, string][] = [\n [\"عائلة تعيش بصمت كامل لأن وحوش تصطاد بالصوت\", \"A Quiet Place\"],\n [\"مسلسل ألماني عن سفر بالزمن فيه كهف وأطفال مفقودين\", \"Dark\"],\n [\"رجل يدخل بالأحلام عشان يزرع فكرة بعقل واحد\", \"Inception\"],\n [\"مسلسل كوري ناس مفلسين يلعبون ألعاب أطفال والخسران يموت\", \"Squid Game\"],\n [\"فلم كرتون عن رجل عجوز يربط بيته ببالونات ويطير\", \"Up\"],\n [\"فلم عن رجل يعيش نفس اليوم مرة ورا مرة\", \"Groundhog\"],\n];\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n const i = Number(url.searchParams.get(\"i\") || \"0\");\n const rows: any[] = [];\n for (const [q, expect] of CASES.slice(i, i + 2)) {\n try {\n const r = await findByKeywords(q);\n const titles = (r.items || []).map((x: any) => x.t);\n rows.push({\n q, expect,\n rankAuto: titles.findIndex((t: string) => new RegExp(expect, \"i\").test(t || \"\")),\n chips: r.debug.chosen,\n top: (r.items || []).slice(0, 4).map((x: any) => x.t + \" [\" + x.hit + \"·\" + x.idf + \"] \" + (x.matched || []).join(\"،\")),\n ms: r.debug.ms,\n });\n } catch (e) { rows.push({ q, err: String(e).slice(0, 200) }); }\n }\n return Response.json(rows);\n}\n",[],[487],"ab935c8a-706c-46dc-9d45-8bda12b6ee1e","kw100.ts","435eeb0e-b0c5-11f1-88f0-1607ee4eb77e",152,["D",1789449671042],"/* تطوير ١: توسيع الوسوم من ٢٥ إلى ١٠٠ — الوسوم المصدر الوحيد المثبَت نفعه */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\n\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; await fn(items[k]); }\n }));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1000\");\n\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = all.slice(start, start + count).filter(r => r.imdb && !r.kw100);\n\n const groups: any[][] = [];\n for (let i = 0; i < slice.length; i += 5) groups.push(slice.slice(i, i + 5));\n\n const t0 = Date.now();\n let got = 0, sum = 0;\n await pool(groups, 8, async (g: any[]) => {\n const alias = g.map((r, i) => `t${i}: title(id: ${JSON.stringify(r.imdb)}) {\n keywords(first: 100) { edges { node { text } } } }`).join(\" \");\n const res = await gql(`query { ${alias} }`);\n g.forEach((r, i) => {\n const kws = ((res?.data?.[`t${i}`]?.keywords?.edges) || []).map((e: any) => e.node?.text).filter(Boolean);\n if (kws.length) { r.kw100 = kws.join(\", \").slice(0, 4000); got++; sum += kws.length; }\n });\n });\n\n await blob.setJSON(\"corpus_v2\", all);\n const done = all.filter(r => r.kw100).length;\n return Response.json({\n عولج: slice.length, \"نجح\": got, \"متوسط عدد الوسوم\": got ? Math.round(sum / got) : 0,\n ms: Date.now() - t0, \"إجمالي من لديه وسوم موسّعة\": done, الإجمالي: all.length,\n التالي: start + count,\n });\n}\n",[],[495],"06ec6f4d-dcd0-47f0-be91-831c4343ece6","jtargets.ts","c1633426-b0be-11f1-b1eb-1607ee4eb77e",124,["D",1789446875967],"/* تضمين موجَّه: يضمن حضور كل عناوين الاختبار بالعيّنة مهما كان موقعها بالقاعدة */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { docText, embed } from \"./jembed.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst DIM = 512;\nfunction packVecs(vecs: number[][]) {\n const f = new Float32Array(vecs.length * DIM);\n vecs.forEach((v, i) => f.set(v, i * DIM));\n const b = new Uint8Array(f.buffer);\n let bin = \"\";\n for (let i = 0; i < b.length; i += 8192) bin += String.fromCharCode(...b.subarray(i, i + 8192));\n return btoa(bin);\n}\n\nexport default async function (req: Request): Promise {\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const norm = (s: string) => String(s || \"\").toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n /* مواقع عناوين الاختبار داخل القاعدة */\n const wanted = CASES.map(c => norm(c[1]));\n const idx: number[] = [];\n const found: string[] = [];\n C.forEach((r, i) => {\n const t = norm(r.title);\n if (wanted.some(w => t === w || t.startsWith(w + \" \") || w.startsWith(t))) { idx.push(i); found.push(r.title + \" @\" + i); }\n });\n\n const out: any = { \"عناوين الاختبار بالقاعدة\": found.length, مواقع: found.slice(0, 60) };\n if (new URL(req.url).searchParams.get(\"dry\")) return Response.json(out);\n\n /* نضمّن هذه المواضع فقط (الباقي مضمَّن تسلسلياً بمسار آخر) */\n const slice = idx.map(i => C[i]);\n const B = 64;\n const all: number[][] = [];\n let tokens = 0;\n for (let i = 0; i < slice.length; i += B) {\n const e = await embed(slice.slice(i, i + B).map(docText), \"retrieval.passage\");\n all.push(...e.vecs); tokens += e.tokens;\n }\n await blob.setJSON(\"vecs_targets\", {\n n: all.length, dim: DIM,\n ids: slice.map(r => r.kind + \":\" + r.id),\n titles: slice.map(r => r.title),\n b64: packVecs(all),\n });\n out.مضمَّن = all.length; out.رموز = tokens;\n return Response.json(out);\n}\n",[],[503],"bede3109-a389-41bc-9677-06b5e491f586","jloop.ts","fb06ea42-b0be-11f1-b636-1607ee4eb77e",130,["D",1789447269320],"/* تضمين متواصل بإيقاع يحترم الحدّ المقاس (١٠٠ ألف رمز/دقيقة):\n دفعة ٣٢ نصاً (~١٠ آلاف رمز) كل ٧ ثوانٍ، ويستمر بالخلفية بعد انتهاء مهلة الطلب. */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { docText, embed } from \"./jembed.ts\";\n\nconst DIM = 512;\nfunction pack(vecs: number[][]) {\n const f = new Float32Array(vecs.length * DIM);\n vecs.forEach((v, i) => f.set(v, i * DIM));\n const b = new Uint8Array(f.buffer);\n let bin = \"\";\n for (let i = 0; i < b.length; i += 8192) bin += String.fromCharCode(...b.subarray(i, i + 8192));\n return btoa(bin);\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"1600\");\n const budget = Number(u.searchParams.get(\"sec\") || \"20\") * 1000;\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const state: any = (await blob.getJSON(\"embed_state\")) || { done: {} };\n const t0 = Date.now();\n const log: string[] = [];\n let tokens = 0;\n\n const B = 32;\n for (let s = from; s < Math.min(to, C.length); s += B) {\n if (Date.now() - t0 > budget) { log.push(\"انتهت المهلة عند \" + s); break; }\n const key = \"v_\" + s;\n if (state.done[key]) continue;\n const slice = C.slice(s, s + B);\n try {\n const e = await embed(slice.map(docText), \"retrieval.passage\");\n tokens += e.tokens;\n await blob.setJSON(key, {\n start: s, n: e.vecs.length, dim: DIM,\n ids: slice.map(r => r.kind + \":\" + r.id),\n titles: slice.map(r => r.title),\n b64: pack(e.vecs),\n });\n state.done[key] = e.vecs.length;\n await blob.setJSON(\"embed_state\", state);\n await new Promise(r => setTimeout(r, 1200)); /* تهدئة خفيفة — المهلة القصيرة تحمي من الحدّ */\n } catch (err) {\n log.push(s + \": \" + String(err).slice(0, 120));\n await new Promise(r => setTimeout(r, 20000)); /* تهدئة أطول عند الرفض */\n }\n }\n\n const total = Object.values(state.done).reduce((a: any, b: any) => a + b, 0);\n return Response.json({ مضمَّن_إجمالاً: total, رموز_هذه_الجولة: tokens, log, ثوانٍ: Math.round((Date.now() - t0) / 1000) });\n}\n",[],[511],"38a884af-6d48-4411-a249-51361decfbfb","jinatest.ts","29531688-b0be-11f1-9904-1607ee4eb77e",162,["D",1789483697031],"/* فحص حالة المفتاح بعد الاستهلاك */\nexport default async function (req: Request): Promise {\n const k = Deno.env.get(\"JINA_KEY2\") || \"\";\n const out: any = {};\n try {\n const r = await fetch(\"https://embeddings-dashboard-api.jina.ai/api/v1/api_key/user?api_key=\" + encodeURIComponent(k),\n { signal: AbortSignal.timeout(20000) });\n out.حساب = await r.json();\n } catch (e) { out.err = String(e).slice(0, 150); }\n\n /* اختبار مباشر لواجهة إعادة الترتيب */\n try {\n const r = await fetch(\"https://api.jina.ai/v1/rerank\", {\n method: \"POST\",\n headers: { \"content-type\": \"application/json\", authorization: \"Bearer \" + k },\n body: JSON.stringify({\n model: \"jina-reranker-v2-base-multilingual\",\n query: \"عائلة تعيش بصمت لأن وحوش تصطاد بالصوت\",\n documents: [\"A Quiet Place: a family lives in silence hiding from creatures that hunt by sound\",\n \"Titanic: a ship hits an iceberg and sinks\"],\n top_n: 2,\n }),\n signal: AbortSignal.timeout(25000),\n });\n const t = await r.text();\n out.اختبار = { status: r.status, body: t.slice(0, 400) };\n } catch (e) { out.اختبار = \"ERR \" + String(e).slice(0, 150); }\n return Response.json(out);\n}\n",[],[519],"de0dbed0-ed0b-4207-9ff2-58658dab6f8f","jembed.ts","acbdf8d0-b0be-11f1-b546-1607ee4eb77e",125,["D",1789446901219],"/* تضمين دلالي عبر Jina — نص مختصر مركّز لتوفير الرموز:\n العنوان + الوسوم + أول ٦٠٠ حرف من الملخّصات (الجوهر عادةً بالبداية). */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst KEY = () => Deno.env.get(\"JINA_KEY2\") || Deno.env.get(\"JINA_KEY\") || \"\";\nconst DIM = 512;\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport function docText(r: any) {\n const s = splitExtra(r.extra);\n return [\n String(r.title || \"\"),\n s.kw ? \"Keywords: \" + s.kw.slice(0, 400) : \"\",\n String(r.ov || \"\").slice(0, 300),\n (r.plots || []).join(\" \").slice(0, 600),\n ].filter(Boolean).join(\". \").slice(0, 1400);\n}\n\nexport async function embed(texts: string[], task: \"retrieval.passage\" | \"retrieval.query\") {\n const r = await fetch(\"https://api.jina.ai/v1/embeddings\", {\n method: \"POST\",\n headers: { \"content-type\": \"application/json\", authorization: \"Bearer \" + KEY() },\n body: JSON.stringify({ model: \"jina-embeddings-v3\", task, dimensions: DIM, input: texts }),\n signal: AbortSignal.timeout(50000),\n });\n const j = await r.json();\n if (!j?.data) throw new Error(JSON.stringify(j).slice(0, 200));\n return { vecs: j.data.map((d: any) => d.embedding as number[]), tokens: j?.usage?.total_tokens || 0 };\n}\n\n/* تخزين المتجهات: Float32 مضغوطة كـbase64 — أصغر بكثير من JSON */\nfunction packVecs(vecs: number[][]) {\n const f = new Float32Array(vecs.length * DIM);\n vecs.forEach((v, i) => f.set(v, i * DIM));\n let bin = \"\";\n const b = new Uint8Array(f.buffer);\n for (let i = 0; i < b.length; i += 8192) bin += String.fromCharCode(...b.subarray(i, i + 8192));\n return btoa(bin);\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"500\");\n const out: any = { start, count };\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n /* الأعمال مرتّبة بالأهمية، ونضمن حضور كل حالات الاختبار ضمن العيّنة */\n const slice = C.slice(start, start + count);\n if (!slice.length) return Response.json({ done: true, total: C.length });\n\n const B = 64;\n const all: number[][] = [];\n let tokens = 0;\n const t0 = Date.now();\n for (let i = 0; i < slice.length; i += B) {\n const texts = slice.slice(i, i + B).map(docText);\n try {\n const e = await embed(texts, \"retrieval.passage\");\n all.push(...e.vecs); tokens += e.tokens;\n /* ⚠️ حدّ مقاس: ١٠٠ ألف رمز/دقيقة. الدفعة ~٢٠ ألف رمز، فنهدّئ ١٣ث بين الدفعات */\n if (i + B < slice.length) await new Promise(r => setTimeout(r, 13000));\n } catch (err) {\n out.error = String(err).slice(0, 200);\n break;\n }\n }\n out.مضمَّن = all.length;\n out.رموز = tokens;\n out.ms = Date.now() - t0;\n\n if (all.length) {\n const key = \"vecs_\" + start;\n await blob.setJSON(key, {\n start, n: all.length, dim: DIM,\n ids: slice.slice(0, all.length).map(r => r.kind + \":\" + r.id),\n titles: slice.slice(0, all.length).map(r => r.title),\n b64: packVecs(all),\n });\n out.مخزَّن = key;\n }\n out.التالي = start + all.length;\n return Response.json(out);\n}\n",[],[527],"2d4b8b0f-b05d-4995-87b5-f4a483067f24","indexP.ts","93f57f30-b131-11f1-ac7d-1607ee4eb77e",193,["D",1789497447061],"/* فهرس بخمسة أعمدة: عنوان · ملخّصات · وسوم · طاقم — الطاقم عمود مستقل بوزن خاص */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"2300\");\n const out: any = {};\n\n if (start === 0) {\n /* ⚠️ عمود عربي منفصل: العناوين والملخّصات العربية تُبحث بالعربي مباشرة بلا ترجمة.\n والمُقسِّم unicode61 يدعم العربية، بخلاف porter المخصَّص للإنجليزية. */\n await sqlite.execute(`DROP TABLE IF EXISTS plotP`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plotP USING fts5(\n title, plots, kw, people, ar, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED,\n tokenize='unicode61 remove_diacritics 2')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n let withPeople = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 30) {\n await sqlite.batch(slice.slice(i, i + 30).map(r => {\n const s = splitExtra(r.extra);\n const ppl = String(r.people || \"\");\n if (ppl) withPeople++;\n /* العمود العربي: العنوان الأصلي + ملخّص TMDB العربي إن وُجد */\n const ar = [String(r.titleAr || \"\"), String(r.ovAr || \"\")].filter(Boolean).join(\" \\n \");\n return {\n sql: `INSERT INTO plotP (title, plots, kw, people, ar, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [s.kw, s.tag].filter(Boolean).join(\" \\n \"),\n ppl, ar,\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plotP`);\n out.inserted = slice.length; out[\"لديه طاقم\"] = withPeople; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[535],"df0efdea-581f-4cae-b286-04ccf94a6a91","indexC.ts","30c81e44-b137-11f1-a4e2-1607ee4eb77e",210,["D",1789499034038],"/* فهرس بعمود النقد — للقياس المقارن على نطاق الأعمال التي لها نقد */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"2400\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plotC`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plotC USING fts5(\n title, plots, kw, people, ar, ad,\n kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED,\n tokenize='unicode61 remove_diacritics 2')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n let withCrit = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 25) {\n await sqlite.batch(slice.slice(i, i + 25).map(r => {\n const s = splitExtra(r.extra);\n const crit = String(r.ad || \"\");\n if (crit) withCrit++;\n return {\n sql: `INSERT INTO plotC (title, plots, kw, people, ar, ad, kind, year, tmdbid)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [s.kw, s.tag].filter(Boolean).join(\" \\n \"),\n String(r.people || \"\"),\n [String(r.titleAr || \"\"), String(r.ovAr || \"\")].filter(Boolean).join(\" \\n \"),\n crit.slice(0, 9000),\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plotC`);\n out.inserted = slice.length; out[\"لديه نقد\"] = withCrit; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[543],"086cd48d-e71a-49f4-8001-3ff8fa2cdb8a","indexA.ts","b8eedc44-b0c5-11f1-acd6-1607ee4eb77e",156,["D",1789449868263],"/* فهرس بحلقات TVmaze كعمود رابع — للقياس قبل إكمال الإثراء */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"2400\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plotA`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plotA USING fts5(\n title, plots, kw, eps, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n let withEps = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 25) {\n await sqlite.batch(slice.slice(i, i + 25).map(r => {\n const s = splitExtra(r.extra);\n const eps = String(r.eps || \"\");\n if (eps) withEps++;\n return {\n sql: `INSERT INTO plotA (title, plots, kw, eps, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [s.kw, s.tag].filter(Boolean).join(\" \\n \"),\n eps.slice(0, 7000),\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plotA`);\n out.inserted = slice.length; out[\"لديه حلقات\"] = withEps; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[551],"3d7cf1be-b9ae-4897-bd71-aaef01fe0fc7","index9.ts","56c77f6c-b0c5-11f1-9ee0-1607ee4eb77e",153,["D",1789449703602],"/* فهرس بالوسوم الموسّعة (٨١ وسماً بدل ٢٥) — للمقارنة المباشرة */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1200\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plot9`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plot9 USING fts5(\n title, plots, kw, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n /* المقارنة عادلة فقط على الأعمال التي أُثريت — أول ١٠٠٠ */\n const slice = C.slice(start, start + count);\n let wide = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 30) {\n await sqlite.batch(slice.slice(i, i + 30).map(r => {\n const s = splitExtra(r.extra);\n const kw = r.kw100 || s.kw; /* الموسّع إن وُجد، وإلا القديم */\n if (r.kw100) wide++;\n return {\n sql: `INSERT INTO plot9 (title, plots, kw, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [kw, s.tag].filter(Boolean).join(\" \\n \"),\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plot9`);\n out.inserted = slice.length; out[\"بوسوم موسّعة\"] = wide; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[559],"77e96226-84db-421d-b399-56b81e6af68e","index8.ts","408eee94-b0c3-11f1-a9ca-1607ee4eb77e",149,["D",1789448807326],"/* الفهرس رباعي الأعمدة — فكرة الخلطة: كل مصدر عمود بوزن مستقل\n ١ العنوان · ٢ ملخّصات IMDb · ٣ الوسوم · ٤ حبكة ويكيبيديا (CMU) + حلقات TVmaze */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"2500\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plot8`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plot8 USING fts5(\n title, plots, kw, story,\n kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n let withStory = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 25) {\n await sqlite.batch(slice.slice(i, i + 25).map(r => {\n const s = splitExtra(r.extra);\n /* العمود الرابع: حبكة ويكيبيديا + ملخّصات الحلقات — المصدران الجديدان */\n const story = [String(r.cmu || \"\"), String(r.ext2 || \"\")].filter(Boolean).join(\" \\n \").slice(0, 9000);\n if (story) withStory++;\n return {\n sql: `INSERT INTO plot8 (title, plots, kw, story, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [s.kw, s.tag].filter(Boolean).join(\" \\n \"),\n story,\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plot8`);\n out.inserted = slice.length; out[\"لديه حبكة\"] = withStory; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[567],"138638d2-edbf-4e39-a85e-b9529f624936","index6.ts","3df4bdb8-b081-11f1-99ef-1607ee4eb77e",111,["D",1789420456174],"/* الحفر في تطبيع الطول: نحذف الكلمات الوظيفية **عند الفهرسة** لا عند البحث.\n الفرق جوهري: الحذف عند البحث يفقد السياق، أما الحذف عند الفهرسة فيمنع\n النصوص الطويلة من تكديس درجات من \"the/and/to\" — بلا فقد أي كلمة مضمون. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been being am \" +\n \"about after before while during into onto over under out up down off through across against between among \" +\n \"who whom whose that this these those which what when where how why there here \" +\n \"he she it they them him her his hers its their theirs we us our ours you your i me my mine \" +\n \"not no nor none any some all both each few more most other such only own same so than too very \" +\n \"can will just should now does do did done has have had having would could may might must \" +\n \"also then thus however therefore because since until though although if else \" +\n \"one two three s t d ll re ve m\").split(\" \"));\n\nconst strip = (s: string) =>\n (String(s || \"\").toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => !STOP.has(w) && w.length > 1).join(\" \");\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1800\");\n const cap = Number(u.searchParams.get(\"cap\") || \"0\"); /* 0 = بلا تقصير */\n const table = cap ? \"plot7\" : \"plot6\";\n const out: any = { table, cap };\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS ${table}`);\n await sqlite.execute(`CREATE VIRTUAL TABLE ${table} USING fts5(\n title, plots, kw, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 40) {\n await sqlite.batch(slice.slice(i, i + 40).map(r => {\n const s = splitExtra(r.extra);\n let body = strip([String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \"));\n if (cap) body = body.split(\" \").slice(0, cap).join(\" \");\n return {\n sql: `INSERT INTO ${table} (title, plots, kw, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"), body, strip([s.kw, s.tag].join(\" \")),\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM ${table}`);\n out.inserted = slice.length; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[575],"7e58dd2d-ec82-4d8b-b23f-680a1b446a8b","index5.ts","092ea8b0-b080-11f1-b300-1607ee4eb77e",117,["D",1789421774697],"/* الحل الجذري: فهرس واحد بعمودين منفصلين (ملخّصات · وسوم).\n FTS5 يدعم وزناً مستقلاً لكل عمود داخل bm25() — فلا دمج رتب ولا فقدان درجات،\n ونضبط الوزن بالقياس بدل الاختيار بين فهرسين. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) return { kw: p[0], tag: p[3] || \"\" };\n return { kw: \"\", tag: \"\" };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1800\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plot5`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plot5 USING fts5(\n title, plots, kw, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED,\n tokenize='porter unicode61')`);\n out.created = true;\n }\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 40) {\n await sqlite.batch(slice.slice(i, i + 40).map(r => {\n const s = splitExtra(r.extra);\n return {\n sql: `INSERT INTO plot5 (title, plots, kw, kind, year, tmdbid) VALUES (?, ?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \")].join(\" \\n \"),\n [s.kw, s.tag].filter(Boolean).join(\" \\n \"),\n String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plot5`);\n out.inserted = slice.length; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[583],"53a8abe7-e4fe-4b17-9f5f-59d0cb8c1041","index4.ts","dc0a33b2-b076-11f1-8592-1607ee4eb77e",101,["D",1789415996938],"/* فهرس رابع: ملخّصات + وسوم + شعارات فقط — بلا اقتباسات ولا طرائف (مصدر التشويش المقاس) */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\n/* extra بُني هكذا: [kw, kw, kw, taglines, quotes, trivia].join(\" \\n \")\n فالمقاطع الثلاثة الأولى متطابقة (الوسوم) والرابع الشعارات. */\nfunction splitExtra(extra: string) {\n const p = String(extra || \"\").split(\" \\n \");\n if (p.length >= 3 && p[0] === p[1] && p[1] === p[2]) {\n return { kw: p[0], taglines: p[3] || \"\", rest: p.slice(4).join(\" \") };\n }\n return { kw: \"\", taglines: \"\", rest: p.join(\" \") };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1500\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plot4`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plot4 USING fts5(\n title, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, body, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n let withKw = 0;\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 40) {\n await sqlite.batch(slice.slice(i, i + 40).map(r => {\n const s = splitExtra(r.extra);\n if (s.kw) withKw++;\n return {\n sql: `INSERT INTO plot4 (title, kind, year, tmdbid, body) VALUES (?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"), String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\"),\n [String(r.ov || \"\"), (r.plots || []).join(\" \"), s.kw, s.kw, s.taglines].filter(Boolean).join(\" \\n \")],\n };\n }));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plot4`);\n out.inserted = slice.length; out.withKeywords = withKw; out.ms = Date.now() - t;\n out.total = rc.rows[0][0]; out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[591],["D",1789415811270],[],[18],"48ad11c9-a4ec-4cca-945c-ace95fd057b0","index2.ts","7e45a40e-b072-11f1-a522-1607ee4eb77e",88,["D",1789414121637],"/* ينقل القاعدة الجديدة لفهرس FTS5 (جدول منفصل حتى لا نكسر القديم قبل التأكد) */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"1500\");\n const out: any = {};\n\n if (start === 0) {\n await sqlite.execute(`DROP TABLE IF EXISTS plot2`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plot2 USING fts5(\n title, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, body, tokenize='porter unicode61')`);\n out.created = true;\n }\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const slice = C.slice(start, start + count);\n const t = Date.now();\n for (let i = 0; i < slice.length; i += 200) {\n await sqlite.batch(slice.slice(i, i + 200).map(r => ({\n sql: `INSERT INTO plot2 (title, kind, year, tmdbid, body) VALUES (?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"), String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\"),\n String(r.ov || \"\") + \" \" + (r.plots || []).join(\" \")],\n })));\n }\n const rc = await sqlite.execute(`SELECT count(*) FROM plot2`);\n out.inserted = slice.length;\n out.ms = Date.now() - t;\n out.totalInTable = rc.rows[0][0];\n out.corpusTotal = C.length;\n out.next = start + count < C.length ? start + count : null;\n return Response.json(out);\n}\n",[],[602],"d2d683fb-48b0-4ab3-876a-b82e4ff33f87","imdbserve.ts","b5669678-b009-11f1-a5b1-1607ee4eb77e",76,["D",1789369116962],"/* يضمن أعمال الاختبار + يخدم قاعدة IMDb للقياس */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\nconst KEY = \"imdbcorpus_v1\";\n\nconst WANT: [string, string][] = [\n [\"movie\", \"A Quiet Place\"], [\"tv\", \"Dark\"], [\"movie\", \"Inception\"], [\"tv\", \"Squid Game\"],\n [\"tv\", \"Severance\"], [\"movie\", \"The Sixth Sense\"], [\"movie\", \"Up\"], [\"movie\", \"Groundhog Day\"],\n [\"tv\", \"Breaking Bad\"], [\"movie\", \"The Lincoln Lawyer\"], [\"movie\", \"The Matrix\"],\n [\"movie\", \"Interstellar\"], [\"movie\", \"Parasite\"], [\"tv\", \"Chernobyl\"], [\"movie\", \"Titanic\"],\n];\n\nasync function j(u: string) {\n try { const r = await fetch(u, { signal: AbortSignal.timeout(15000) }); return r.ok ? await r.json() : null; }\n catch { return null; }\n}\nasync function plots(id: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, signal: AbortSignal.timeout(20000),\n body: JSON.stringify({ query: `query { title(id: ${JSON.stringify(id)}) {\n plots(first: 6) { edges { node { plotText { plainText } } } } } }` }) });\n const d = await r.json();\n return (d?.data?.title?.plots?.edges || []).map((e: any) => String(e.node?.plotText?.plainText || \"\")).filter(Boolean);\n } catch { return []; }\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const all: any[] = (await blob.getJSON(KEY)) as any[] || [];\n\n if (u.searchParams.get(\"ensure\")) {\n const have = new Set(all.map(r => String(r.title || \"\").toLowerCase()));\n const added: string[] = [];\n for (const [kind, name] of WANT) {\n if (have.has(name.toLowerCase())) continue;\n const s = await j(`${SH}/tmdb/search/${kind}?query=${encodeURIComponent(name)}`);\n const x = s?.results?.[0];\n if (!x) continue;\n const e = await j(`${SH}/tmdb/${kind}/${x.id}/external_ids`);\n if (!e?.imdb_id) continue;\n all.push({\n id: x.id, kind, title: x.title || x.name, imdb: e.imdb_id,\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n plots: await plots(e.imdb_id),\n });\n added.push(name);\n }\n await blob.setJSON(KEY, all);\n return Response.json({ added, total: all.length });\n }\n\n return Response.json(all.map(r => ({\n t: r.title, y: r.year, k: r.kind,\n ov: r.ov,\n p: (r.plots || []).join(\" \\n \"),\n })), { headers: { \"access-control-allow-origin\": \"*\" } });\n}\n",[],[610],"58fcb378-0d8d-4c20-9013-c7fc5129d460","imdbbuild.ts","96cb2102-b009-11f1-ba79-1607ee4eb77e",75,["D",1789369065617],"/* يبني قاعدة قياس مصغَّرة بملخّصات IMDb الكاملة (٦ لكل عمل) */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\nconst KEY = \"imdbcorpus_v1\";\n\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function j(u: string) {\n try { const r = await fetch(u, { signal: AbortSignal.timeout(15000) }); return r.ok ? await r.json() : null; }\n catch { return null; }\n}\n\n/* نسحب ملخّصات ٥ أعمال بنداء واحد عبر الأسماء المستعارة — يقلّل النداءات ٥ أضعاف */\nasync function plotsBatch(ids: string[]) {\n const alias = ids.map((id, i) => `t${i}: title(id: ${JSON.stringify(id)}) {\n titleText { text } plots(first: 6) { edges { node { plotText { plainText } plotType } } } }`).join(\" \");\n const r = await gql(`query { ${alias} }`);\n return ids.map((id, i) => {\n const t = r?.data?.[`t${i}`];\n const texts = (t?.plots?.edges || []).map((e: any) => String(e.node?.plotText?.plainText || \"\")).filter(Boolean);\n return { id, imdbTitle: t?.titleText?.text || \"\", plots: texts };\n });\n}\n\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n const out: R[] = []; let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; out[k] = await fn(items[k]); }\n }));\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"1\");\n const to = Number(u.searchParams.get(\"to\") || \"3\");\n\n let prev: any[] = [];\n try { prev = (await blob.getJSON(KEY)) as any[] || []; } catch { /* أول مرة */ }\n const seen = new Set(prev.map((r: any) => r.kind + \":\" + r.id));\n\n /* ١) نجمع أعمالاً من TMDB مع معرّف IMDb */\n const pages: [string, number][] = [];\n for (let p = from; p <= to; p++) { pages.push([\"movie\", p]); pages.push([\"tv\", p]); }\n const base = (await pool(pages, 6, async ([kind, p]) => {\n const d = await j(`${SH}/tmdb/discover/${kind}?sort_by=popularity.desc&page=${p}&vote_count.gte=100`);\n return (d?.results || []).map((x: any) => ({\n id: x.id, kind, title: x.title || x.name,\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n }));\n })).flat().filter((r: any) => !seen.has(r.kind + \":\" + r.id));\n\n /* ٢) معرّف IMDb لكل عمل */\n const withImdb = await pool(base, 12, async (r: any) => {\n const e = await j(`${SH}/tmdb/${r.kind}/${r.id}/external_ids`);\n r.imdb = e?.imdb_id || null;\n return r;\n });\n const ok = withImdb.filter((r: any) => r.imdb);\n\n /* ٣) ملخّصات IMDb — دفعات من ٥ */\n const groups: any[][] = [];\n for (let i = 0; i < ok.length; i += 5) groups.push(ok.slice(i, i + 5));\n const plotted = await pool(groups, 5, async (g: any[]) => {\n const res = await plotsBatch(g.map(r => r.imdb));\n g.forEach((r, i) => { r.plots = res[i]?.plots || []; });\n return g;\n });\n\n const all = prev.concat(plotted.flat());\n await blob.setJSON(KEY, all);\n const withPlots = all.filter((r: any) => (r.plots || []).length);\n return Response.json({\n added: plotted.flat().length, total: all.length,\n withPlots: withPlots.length,\n avgPlots: +(withPlots.reduce((a: number, r: any) => a + r.plots.length, 0) / (withPlots.length || 1)).toFixed(1),\n });\n}\n",[],[618],"280e4a4f-4566-459c-92bb-8443a1416b83","imdbbench.ts","d212ec36-b009-11f1-b6b8-1607ee4eb77e",77,["D",1789369165061],"/* القياس المقارن داخل Val Town: ملخّص TMDB مقابل ملخّصات IMDb — نفس الخوارزمية بالضبط */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\n\nconst CASES: [string, string][] = [\n [\"عائلة تعيش بصمت كامل لأن وحوش تصطاد بالصوت\", \"A Quiet Place\"],\n [\"مسلسل ألماني عن سفر بالزمن فيه كهف وأطفال مفقودين\", \"Dark\"],\n [\"رجل يدخل بالأحلام عشان يزرع فكرة بعقل واحد\", \"Inception\"],\n [\"مسلسل كوري ناس مفلسين يلعبون ألعاب أطفال والخسران يموت\", \"Squid Game\"],\n [\"مسلسل عن شركة موظفينها يفصلون ذاكرة الشغل عن حياتهم\", \"Severance\"],\n [\"فلم عن ولد صغير يقدر يشوف الأموات\", \"The Sixth Sense\"],\n [\"فلم كرتون عن رجل عجوز يربط بيته ببالونات ويطير\", \"Up\"],\n [\"فلم عن رجل يعيش نفس اليوم مرة ورا مرة\", \"Groundhog Day\"],\n [\"مسلسل عن مدرس كيمياء يصير يصنع مخدرات\", \"Breaking Bad\"],\n [\"فلم عن محامي يشتغل من سيارته\", \"The Lincoln Lawyer\"],\n [\"فلم عن عالم افتراضي والبطل ياخذ حبة حمرا\", \"The Matrix\"],\n [\"فلم عن سفر بالفضاء عبر ثقب دودي عشان ينقذون البشرية\", \"Interstellar\"],\n [\"فلم كوري عن عائلة فقيرة تشتغل عند عائلة غنية\", \"Parasite\"],\n [\"مسلسل عن انفجار مفاعل نووي بالاتحاد السوفيتي\", \"Chernobyl\"],\n [\"فلم عن سفينة تضرب جبل جليد وتغرق\", \"Titanic\"],\n];\n\nconst tok = (s: string) => (s || \"\").toLowerCase().match(/[a-z0-9]+/g) || [];\n\n/* BM25 بسيط ومطابق للمكتبة القياسية بمعاملاتها الافتراضية */\nfunction buildBM25(docs: string[][]) {\n const N = docs.length;\n const df = new Map();\n for (const d of docs) for (const w of new Set(d)) df.set(w, (df.get(w) || 0) + 1);\n const len = docs.map(d => d.length);\n const avg = len.reduce((a, b) => a + b, 0) / N;\n const tf = docs.map(d => { const m = new Map(); for (const w of d) m.set(w, (m.get(w) || 0) + 1); return m; });\n const k1 = 1.5, b = 0.75;\n return (q: string[]) => {\n const s = new Float64Array(N);\n for (const w of q) {\n const n = df.get(w); if (!n) continue;\n const idf = Math.log(1 + (N - n + 0.5) / (n + 0.5));\n for (let i = 0; i < N; i++) {\n const f = tf[i].get(w); if (!f) continue;\n s[i] += idf * (f * (k1 + 1)) / (f + k1 * (1 - b + b * len[i] / avg));\n }\n }\n return s;\n };\n}\n\nexport default async function (req: Request): Promise {\n const C: any[] = (await blob.getJSON(\"imdbcorpus_v1\")) as any[] || [];\n const titles = C.map(r => String(r.title || \"\"));\n\n const bmTMDB = buildBM25(C.map(r => tok(r.ov)));\n const bmIMDB = buildBM25(C.map(r => tok((r.plots || []).join(\" \"))));\n const bmBOTH = buildBM25(C.map(r => tok(r.ov + \" \" + (r.plots || []).join(\" \"))));\n\n const res: any = { corpus: C.length, cases: [] };\n const tally: any = { tmdb: [0, 0, 0], imdb: [0, 0, 0], both: [0, 0, 0] };\n\n for (const [ar, exp] of CASES) {\n let en = ar;\n try {\n const t = await fetch(`${SH}/translate?q=${encodeURIComponent(ar)}&target=en`, { signal: AbortSignal.timeout(15000) })\n .then(r => r.ok ? r.json() : null);\n if (t?.text) en = String(t.text);\n } catch { /* نكمل بالعربي */ }\n const q = tok(en);\n\n const row: any = { exp };\n for (const [name, fn] of [[\"tmdb\", bmTMDB], [\"imdb\", bmIMDB], [\"both\", bmBOTH]] as [string, any][]) {\n const s = fn(q);\n const order = Array.from(s.keys()).sort((a: number, b: number) => s[b] - s[a]);\n const rank = order.findIndex((i: number) => titles[i].toLowerCase() === exp.toLowerCase());\n row[name] = rank;\n row[name + \"Top\"] = titles[order[0]];\n if (rank === 0) tally[name][0]++;\n if (rank >= 0 && rank < 5) tally[name][1]++;\n if (rank >= 0 && rank < 20) tally[name][2]++;\n }\n res.cases.push(row);\n }\n res.summary = {\n \"ملخص TMDB\": `أول ${tally.tmdb[0]}/15 · ضمن٥ ${tally.tmdb[1]}/15 · ضمن٢٠ ${tally.tmdb[2]}/15`,\n \"ملخصات IMDb\": `أول ${tally.imdb[0]}/15 · ضمن٥ ${tally.imdb[1]}/15 · ضمن٢٠ ${tally.imdb[2]}/15`,\n \"الاثنان معاً\": `أول ${tally.both[0]}/15 · ضمن٥ ${tally.both[1]}/15 · ضمن٢٠ ${tally.both[2]}/15`,\n };\n return Response.json(res);\n}\n",[],[626],"52038205-3cbb-440f-b943-96f484fbf6db","fusion.ts","f61c7c70-b07f-11f1-8667-1607ee4eb77e",103,["D",1789419906144],"/* الحل بلا مقايضة: نسأل الفهرسين معاً وندمج بالرتب (RRF).\n العمل الذي يعلو بأي فهرس يرتفع بالنتيجة النهائية — فلا نخسر ما ربحه أيٌّ منهما. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { CASES } from \"./cases.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\n\nasync function run(table: string, terms: string[], lim = 40) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY bm25(${table},1.0,1.0) LIMIT ${lim}`,\n args: [terms.map(w => `\"${w}\"`).join(\" OR \")],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\n/* دمج الرتب: كل قائمة تعطي العمل نقاطاً بحسب موقعه فيها.\n K=60 قيمة قياسية بأبحاث الاسترجاع — تمنع سيطرة المركز الأول المطلقة. */\nfunction rrf(lists: string[][], weights: number[], K = 60) {\n const s = new Map();\n lists.forEach((l, li) => l.forEach((t, i) => {\n s.set(t, (s.get(t) || 0) + weights[li] / (K + i + 1));\n }));\n return [...s.entries()].sort((a, b) => b[1] - a[1]).map(([t]) => t);\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"12\");\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"أصلي\", \"وسوم\", \"دمج ١:١\", \"دمج ٢:١ للأصلي\", \"دمج ١:٢ للوسوم\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n let en = ar;\n if (/[\\u0600-\\u06FF]/.test(ar)) {\n try {\n const t = await fetch(`${SH}/translate?q=${encodeURIComponent(ar)}&target=en`, { signal: AbortSignal.timeout(12000) })\n .then(r => r.ok ? r.json() : null);\n if (t?.text) en = String(t.text);\n } catch { /* نكمل */ }\n }\n const terms = (en.toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => w.length > 1);\n const A = await run(\"plot2\", terms);\n const B = await run(\"plot4\", terms);\n\n const lists: Record = {\n \"أصلي\": A, \"وسوم\": B,\n \"دمج ١:١\": rrf([A, B], [1, 1]),\n \"دمج ٢:١ للأصلي\": rrf([A, B], [2, 1]),\n \"دمج ١:٢ للوسوم\": rrf([A, B], [1, 2]),\n };\n\n const want = norm(exp);\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : \"#\" + (rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" | \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.join(\" | \"),\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[634],"5a5b7fae-d53e-4632-9966-556182d45290","fts.ts","1738d752-b01a-11f1-938c-1607ee4eb77e",83,["D",1789376153020],"/* اختبار معماري حاسم: SQLite FTS5 بدل إعادة بناء الفهرس بكل نداء.\n FTS5 يخزّن الفهرس مرة واحدة ويرتّب بـBM25 داخلياً — نفس الخوارزمية، بلا كلفة بناء. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const out: any = {};\n\n if (u.searchParams.get(\"build\")) {\n let t = Date.now();\n await sqlite.execute(`DROP TABLE IF EXISTS plotfts`);\n await sqlite.execute(`CREATE VIRTUAL TABLE plotfts USING fts5(\n title, kind UNINDEXED, year UNINDEXED, tmdbid UNINDEXED, body, tokenize='porter unicode61')`);\n out.createMs = Date.now() - t;\n\n t = Date.now();\n const C: any[] = (await blob.getJSON(\"imdbcorpus_v1\")) as any[] || [];\n out.loadMs = Date.now() - t;\n\n t = Date.now();\n const B = 200;\n let n = 0;\n for (let i = 0; i < C.length; i += B) {\n const chunk = C.slice(i, i + B);\n const stmts = chunk.map(r => ({\n sql: `INSERT INTO plotfts (title, kind, year, tmdbid, body) VALUES (?, ?, ?, ?, ?)`,\n args: [String(r.title || \"\"), String(r.kind || \"\"), String(r.year || \"\"), String(r.id || \"\"),\n String(r.ov || \"\") + \" \" + (r.plots || []).join(\" \")],\n }));\n await sqlite.batch(stmts);\n n += chunk.length;\n }\n out.insertMs = Date.now() - t;\n out.inserted = n;\n return Response.json(out);\n }\n\n /* بحث */\n const q = u.searchParams.get(\"q\") || \"family lives in complete silence because monsters hunt by sound\";\n /* FTS5 يحتاج تنظيف الرموز الخاصة، ونستعمل OR لأن AND يقصي كثيراً */\n const terms = (q.toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => w.length > 1);\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n\n const t = Date.now();\n const r = await sqlite.execute({\n sql: `SELECT title, kind, year, tmdbid, bm25(plotfts, 1.0, 1.0) AS score\n FROM plotfts WHERE plotfts MATCH ? ORDER BY score LIMIT 10`,\n args: [match],\n });\n out.searchMs = Date.now() - t;\n out.terms = terms.length;\n out.results = (r.rows || []).map((x: any) => `${x[0]} (${x[2]}) ${Number(x[4]).toFixed(2)}`);\n return Response.json(out);\n}\n",[],[642],"aa00b4f2-7ded-40c9-b5ed-1adbbe015460","finalbench.ts","8f92fc08-b076-11f1-8e51-1607ee4eb77e",102,["D",1789416048918],"/* القياس الحاسم الثلاثي: أصلي · مُثرى كامل · وسوم فقط */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { CASES } from \"./cases.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\n\nconst T = [[\"أصلي\", \"plot2\"], [\"مُثرى كامل\", \"plot3\"], [\"وسوم فقط\", \"plot4\"]] as [string, string][];\n\nasync function run(table: string, terms: string[]) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY bm25(${table},1.0,1.0) LIMIT 20`,\n args: [terms.map(w => `\"${w}\"`).join(\" OR \")],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"12\");\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const tally: any = {}; for (const [n] of T) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n let en = ar;\n if (/[\\u0600-\\u06FF]/.test(ar)) {\n try {\n const t = await fetch(`${SH}/translate?q=${encodeURIComponent(ar)}&target=en`, { signal: AbortSignal.timeout(12000) })\n .then(r => r.ok ? r.json() : null);\n if (t?.text) en = String(t.text);\n } catch { /* نكمل */ }\n }\n const terms = (en.toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => w.length > 1);\n const want = norm(exp);\n const parts: string[] = [];\n for (const [name, tbl] of T) {\n const list = await run(tbl, terms);\n const rk = list.findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : \"#\" + (rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" | \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n ترتيب_الأعمدة: \"أصلي | مُثرى كامل | وسوم فقط\",\n النتيجة: Object.fromEntries(T.map(([name]) =>\n [name, `أول ${tally[name][0]}/${n} · ضمن٥ ${tally[name][1]}/${n} · ضمن٢٠ ${tally[name][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[650],"3eda2f21-e465-43f9-8aab-97660e0c4673","final.ts","533ce846-b0cf-11f1-b531-1607ee4eb77e",158,["D",1789453992625],"/* القياس النهائي على القاعدة الكاملة — التركيبة المقترحة مقابل الحالية */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst V: [string, string, string][] = [\n [\"الحالي (plot5)\", \"plot5\", \"bm25(plot5,1,1,1)\"],\n [\"+حلقات ٠٫٥\", \"plotA\", \"bm25(plotA,1,1,1,0.5)\"],\n [\"+حلقات ١\", \"plotA\", \"bm25(plotA,1,1,1,1)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"23\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const tally: any = {}; for (const [n] of V) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const want = norm(exp);\n const parts: string[] = [];\n\n for (const [name, table, rank] of V) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n const list = (r.rows || []).map((x: any) => String(x[0]));\n const rk = list.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.map(v => v[0]).join(\" · \"),\n النتيجة: Object.fromEntries(V.map(([name]) =>\n [name, `أول ${tally[name][0]}/${n} · ضمن٥ ${tally[name][1]}/${n} · ضمن٢٠ ${tally[name][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[658],"88574b29-8173-4725-b5be-86c3de11b969","epsbench.ts","d6d25b82-b0c5-11f1-9042-1607ee4eb77e",157,["D",1789449918413],"/* قياس أثر حلقات TVmaze — على نطاق أول ٢٤٠٠ عمل (٦٩٤ منها بحلقات) */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst V: [string, string][] = [\n [\"بلا حلقات\", \"bm25(plotA,1,1,1,0)\"],\n [\"حلقات وزن ١\", \"bm25(plotA,1,1,1,1)\"],\n [\"حلقات وزن ٠٫٥\", \"bm25(plotA,1,1,1,0.5)\"],\n [\"حلقات وزن ٢\", \"bm25(plotA,1,1,1,2)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const scope = await sqlite.execute(`SELECT title FROM plotA`);\n const allowed = new Set((scope.rows || []).map((x: any) => String(x[0])));\n\n const tally: any = {}; for (const [n] of V) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n let tested = 0;\n\n for (const [ar, exp] of CASES) {\n const want = norm(exp);\n if (![...allowed].some(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); })) continue;\n tested++;\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const parts: string[] = [];\n\n for (const [name, rank] of V) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM plotA WHERE plotA MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n const list = (r.rows || []).map((x: any) => String(x[0]));\n const rk = list.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n return Response.json({\n \"نطاق المقارنة\": allowed.size, \"حالات مختبَرة\": tested,\n أعمدة: V.map(v => v[0]).join(\" · \"),\n النتيجة: Object.fromEntries(V.map(([n]) =>\n [n, `أول ${tally[n][0]}/${tested} · ضمن٥ ${tally[n][1]}/${tested} · ضمن٢٠ ${tally[n][2]}/${tested}`])),\n التفصيل: rows,\n });\n}\n",[],[666],"717473e5-9190-493c-93f9-8ae775a097bd","enrich3.ts","22350326-b0c2-11f1-bd9b-1607ee4eb77e",145,["D",1789448506535],"/* الطبقة المدمجة: ويكيبيديا (أي لغة) + TVmaze (حلقات) — تُضاف لما عندنا من IMDb */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst UA = { \"user-agent\": \"shashti/1.0 (personal media tracker)\" };\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst KEY = \"corpus_v2\";\n\nconst HEADS = \"Plot|Synopsis|Premise|Plot summary|Story|Storyline|Trama|Sinossi|Argumento|Sinopsis|\" +\n \"Intrigue|Handlung|Enredo|Fabuła|Сюжет|あらすじ|줄거리|剧情|劇情|قصة|الحبكة|ملخص|Résumé|Yarn\";\n\nasync function j(u: string) {\n /* ⚠️ حدّ ويكيبيديا مقاس بالتتبّع: أول ٤ طلبات تنجح ثم تُخنَق البقية.\n التهدئة هنا لا بالتوازي — لأن خفض التوازي وحده لم يغيّر شيئاً. */\n for (let attempt = 0; attempt < 3; attempt++) {\n try {\n const r = await fetch(u, { headers: UA, signal: AbortSignal.timeout(12000) });\n if (r.ok) return await r.json();\n if (r.status === 429 || r.status >= 500) { await new Promise(s => setTimeout(s, 1500 * (attempt + 1))); continue; }\n return null;\n } catch { await new Promise(s => setTimeout(s, 900)); }\n }\n return null;\n}\n\nasync function wikiPlot(lang: string, title: string) {\n const d = await j(`https://${lang}.wikipedia.org/w/api.php?action=query&prop=extracts&explaintext=1&format=json&redirects=1&titles=${encodeURIComponent(title)}`);\n const p: any = Object.values(d?.query?.pages || {})[0];\n const t = String(p?.extract || \"\");\n if (t.length < 200) return \"\";\n const m = t.match(new RegExp(`\\\\n==+\\\\s*(${HEADS})\\\\s*==+\\\\n([\\\\s\\\\S]*?)(?=\\\\n==+)`, \"i\"));\n /* لو لا قسم حبكة، نأخذ المقدمة — غالباً تلخّص القصة */\n return (m ? m[2] : t.slice(0, 1200)).trim().slice(0, 4000);\n}\n\n/* البحث أدق من تخمين صيغة العنوان — «(2025 film)» فشلت مراراً بالقياس */\nasync function wikiFind(lang: string, q: string) {\n const s = await j(`https://${lang}.wikipedia.org/w/api.php?action=query&list=search&srlimit=3&format=json&srsearch=${encodeURIComponent(q)}`);\n for (const h of (s?.query?.search || []).map((x: any) => x.title)) {\n const t = await wikiPlot(lang, h);\n if (t.length > 250) return t;\n }\n return \"\";\n}\n\nasync function tvmaze(name: string) {\n const d = await j(`https://api.tvmaze.com/singlesearch/shows?q=${encodeURIComponent(name)}&embed=episodes`);\n if (!d?.id) return \"\";\n const eps = (d?._embedded?.episodes || []) as any[];\n const txt = eps.map(e => String(e.name || \"\") + \": \" + String(e.summary || \"\").replace(/<[^>]+>/g, \" \"))\n .join(\" \\n \").replace(/\\s+/g, \" \").trim();\n return txt.slice(0, 6000);\n}\n\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; await fn(items[k]); }\n }));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"120\");\n\n const all: any[] = (await blob.getJSON(KEY)) as any[] || [];\n const slice = all.slice(start, start + count).filter(r => !r.ext2);\n const t0 = Date.now();\n let wikiOk = 0, tvOk = 0;\n const probe: any[] = [];\n\n await pool(slice, 3, async (r: any) => {\n let wiki = \"\";\n const dbg: any = { title: r.title };\n try {\n const qy = r.title + \" \" + (r.kind === \"tv\" ? \"TV series\" : \"film\") + \" \" + (r.year || \"\");\n dbg.query = qy;\n wiki = await wikiFind(\"en\", qy);\n dbg.enLen = wiki.length;\n if (!wiki) {\n const d = await j(`${SH}/tmdb/${r.kind}/${r.id}`);\n const lang = String(d?.original_language || \"\");\n const orig = String(d?.original_title || d?.original_name || r.title);\n dbg.lang = lang;\n if (lang && lang !== \"en\") { wiki = await wikiFind(lang, orig); dbg.langLen = wiki.length; }\n }\n } catch (e) { dbg.err = String(e).slice(0, 160); }\n if (probe.length < 6) probe.push(dbg);\n await new Promise(s => setTimeout(s, 400)); /* فاصل بين الأعمال يمنع الخنق */\n\n let eps = \"\";\n if (r.kind === \"tv\") { try { eps = await tvmaze(r.title); } catch { /* نكمل */ } }\n\n if (wiki) wikiOk++;\n if (eps) tvOk++;\n r.ext2 = [wiki, eps].filter(Boolean).join(\" \\n \").slice(0, 9000);\n });\n\n await blob.setJSON(KEY, all);\n const done = all.filter(r => r.ext2 !== undefined).length;\n const withText = all.filter(r => r.ext2).length;\n return Response.json({\n عولج: slice.length, ويكيبيديا: wikiOk, tvmaze: tvOk,\n ms: Date.now() - t0,\n \"أُنجز من القاعدة\": done, \"لديه نص إضافي\": withText, الإجمالي: all.length,\n تتبّع: probe,\n التالي: start + count,\n });\n}\n",[],[674],"58e27435-2674-4dbb-93c7-754ee1424c9c","enrich2.ts","31f58a3e-b076-11f1-b8da-1607ee4eb77e",97,["D",1789415711556],"/* إثراء القاعدة بالمصادر الأربعة المكتشَفة: وسوم · اقتباسات · شعارات · معلومات طريفة */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\nconst KEY = \"corpus_v2\";\n\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; await fn(items[k]); }\n }));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"700\");\n\n const all: any[] = (await blob.getJSON(KEY)) as any[] || [];\n const slice = all.slice(start, start + count).filter(r => r.imdb && !r.extra);\n\n const groups: any[][] = [];\n for (let i = 0; i < slice.length; i += 5) groups.push(slice.slice(i, i + 5));\n\n const t0 = Date.now();\n await pool(groups, 6, async (g: any[]) => {\n const alias = g.map((r, i) => `t${i}: title(id: ${JSON.stringify(r.imdb)}) {\n taglines(first: 4) { edges { node { text } } }\n quotes(first: 10) { edges { node { lines { text } } } }\n trivia(first: 6) { edges { node { text { plainText } } } }\n keywords(first: 25) { edges { node { text } } } }`).join(\" \");\n const res = await gql(`query { ${alias} }`);\n g.forEach((r, i) => {\n const t = res?.data?.[`t${i}`];\n if (!t) return;\n const tags = (t.taglines?.edges || []).map((e: any) => e.node?.text).filter(Boolean);\n const qs = (t.quotes?.edges || []).map((e: any) =>\n (e.node?.lines || []).map((l: any) => l.text).filter(Boolean).join(\" \")).filter(Boolean);\n const tr = (t.trivia?.edges || []).map((e: any) => e.node?.text?.plainText).filter(Boolean);\n const kw = (t.keywords?.edges || []).map((e: any) => e.node?.text).filter(Boolean);\n /* الوسوم تتكرر ثلاثاً: كلماتها أدق تمييزاً من نثر الاقتباسات، والتكرار يرفع وزنها بـBM25 */\n r.extra = [kw.join(\", \"), kw.join(\", \"), kw.join(\", \"), tags.join(\" \"), qs.join(\" \"), tr.join(\" \")]\n .filter(Boolean).join(\" \\n \").slice(0, 6000);\n });\n });\n\n await blob.setJSON(KEY, all);\n const done = all.filter(r => r.extra).length;\n return Response.json({\n معالَج: slice.length, ms: Date.now() - t0,\n \"لديه إثراء\": done, الإجمالي: all.length,\n التالي: start + count < all.length ? start + count : null,\n });\n}\n",[],[682],"4bfedf98-1822-4183-8af2-03e15ea129c7","engine.ts","5a8211d8-af9b-11f1-9a30-1607ee4eb77e",["D",1789322565717],"/* ابحث بالوصف — الأنبوب المجاني بالكامل (نسخة ٢: تصويت + تحقّق + عدّ تكرار)\n وصف → /translate (خادم شاشتي) → ٣ بحوث متوازية بـSearXNG (أوراكل) →\n تصويت + عدّ تكرار بالنصوص → TMDB → تحقّق من الملخص → بطاقات.\n صفر ذكاء اصطناعي · صفر حصة · صفر طرف ثالث مدفوع. */\n\nconst SHASHTI = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst SEARX = \"http://145.241.108.185:8888/search\";\n\nconst NOISE = [\n \"imdb\", \"rotten tomatoes\", \"letterboxd\", \"netflix\", \"wikipedia\", \"youtube\",\n \"metacritic\", \"justwatch\", \"prime video\", \"reddit\", \"trailer\", \"official trailer\",\n \"movie review\", \"film review\", \"watch online\", \"cast and crew\", \"plot summary\",\n \"the movie database\", \"tmdb\", \"moviefone\", \"fandom\", \"quotes\", \"ending explained\",\n \"best movies\", \"best tv shows\", \"list of films\", \"similar movies\",\n];\n\n/* كلمات لا تحمل معنى تمييزياً — تُستبعد من التحقّق بالملخص (ب) */\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been \" +\n \"his her its their this that these those who whom which what when where how why \" +\n \"about after before during while into onto over under between out up down off \" +\n \"he she it they them him we you i my your our movie film series show tv season episode \" +\n \"one two three new old man woman boy girl people story about make makes made get gets \" +\n \"must can will would could should has have had do does did not no all some any more most\").split(\" \"));\n\nfunction words(s: string): string[] {\n return String(s || \"\").toLowerCase().replace(/[^a-z0-9\\u0600-\\u06FF]+/g, \" \")\n .split(\" \").filter(w => w.length > 2 && !STOP.has(w));\n}\n\nfunction cleanTitle(raw: string): string {\n let t = String(raw || \"\");\n t = t.split(/\\s+[-–|:]\\s+/)[0];\n t = t.replace(/\$(?:TV\\s+(?:Series|Mini[- ]?Series|Movie)\\s*)?\\d{4}(?:[–-]\\s*\\d{0,4})?\$/gi, \"\");\n t = t.replace(/\$\\s*(?:TV\\s+Series|Film|Movie|مسلسل|فيلم)\\s*\$/gi, \"\");\n t = t.replace(/[\"'“”‘’]/g, \"\").replace(/\\s+/g, \" \").trim();\n t = t.replace(/^(?:watch|stream|أشاهد|شاهد)\\s+/i, \"\").trim();\n return t;\n}\n\nfunction isNoise(t: string): boolean {\n const low = t.toLowerCase();\n if (low.length < 2 || low.length > 60) return true;\n return NOISE.some(n => low === n || low.startsWith(n + \" \") || low.endsWith(\" \" + n));\n}\n\ntype Cand = {\n key: string; imdb?: string; tmdb?: string; tmdbType?: string; title?: string;\n score: number; queries: Set; mentions: number; why: string[];\n};\n\nasync function searx(q: string) {\n const u = SEARX + \"?q=\" + encodeURIComponent(q) + \"&format=json\";\n const r = await fetch(u, { signal: AbortSignal.timeout(22000) });\n if (!r.ok) throw new Error(\"searx \" + r.status);\n const j = await r.json();\n return (j?.results || []) as any[];\n}\n\n/* ═══ (أ) تصويت: نجمع مرشَّحي عدة بحوث بمخزن واحد ═══ */\nfunction harvest(map: Map, results: any[], qTag: string, qWeight: number) {\n const bump = (key: string, w: number, patch: Partial, why: string) => {\n let c = map.get(key);\n if (!c) { c = { key, score: 0, queries: new Set(), mentions: 0, why: [] }; map.set(key, c); }\n for (const k in patch) if ((patch as any)[k] != null && (c as any)[k] == null) (c as any)[k] = (patch as any)[k];\n c.score += w * qWeight;\n c.queries.add(qTag);\n if (c.why.length < 3) c.why.push(why);\n };\n\n results.slice(0, 20).forEach((res, i) => {\n const url = String(res?.url || \"\");\n const title = String(res?.title || \"\");\n const w = 1 / (i + 1);\n\n const tt = url.match(/imdb\\.com\\/title\\/(tt\\d+)/i);\n if (tt) { bump(\"imdb:\" + tt[1], w * 3, { imdb: tt[1], title: cleanTitle(title) }, title.slice(0, 90)); return; }\n\n const tm = url.match(/themoviedb\\.org\\/(movie|tv)\\/(\\d+)/i);\n if (tm) { bump(\"tmdb:\" + tm[1] + \":\" + tm[2], w * 3, { tmdb: tm[2], tmdbType: tm[1].toLowerCase(), title: cleanTitle(title) }, title.slice(0, 90)); return; }\n\n if (/rottentomatoes\\.com|letterboxd\\.com|netflix\\.com|wikipedia\\.org|metacritic\\.com|justwatch\\.com|thetvdb\\.com|trakt\\.tv/i.test(url)) {\n const t = cleanTitle(title);\n if (t && !isNoise(t)) bump(\"name:\" + t.toLowerCase(), w * 2, { title: t }, title.slice(0, 90));\n }\n });\n}\n\n/* ═══ (ج) عدّ تكرار اسم العمل داخل نصوص كل النتائج ═══\n العمل الصحيح يُذكر بعناوين ومقتطفات كثيرة، لا برابط واحد فقط. */\nfunction countMentions(cands: Cand[], allText: string) {\n const hay = allText.toLowerCase();\n for (const c of cands) {\n const t = (c.title || \"\").toLowerCase().trim();\n if (t.length < 3) { c.mentions = 0; continue; }\n let n = 0, from = 0;\n while (true) {\n const at = hay.indexOf(t, from);\n if (at === -1) break;\n n++; from = at + t.length;\n if (n > 40) break;\n }\n c.mentions = n;\n }\n}\n\nasync function tmdbFind(imdb: string) {\n const r = await fetch(SHASHTI + \"/tmdb/find/\" + imdb + \"?external_source=imdb_id\", { signal: AbortSignal.timeout(15000) });\n if (!r.ok) return null;\n const j = await r.json();\n const mv = j?.movie_results?.[0], tv = j?.tv_results?.[0];\n if (mv) return { ...mv, _isMovie: true };\n if (tv) return { ...tv, _isMovie: false };\n return null;\n}\n\nasync function tmdbSearch(name: string) {\n const [m, t] = await Promise.all([\n fetch(SHASHTI + \"/tmdb/search/movie?query=\" + encodeURIComponent(name), { signal: AbortSignal.timeout(15000) }).then(r => r.ok ? r.json() : null).catch(() => null),\n fetch(SHASHTI + \"/tmdb/search/tv?query=\" + encodeURIComponent(name), { signal: AbortSignal.timeout(15000) }).then(r => r.ok ? r.json() : null).catch(() => null),\n ]);\n const norm = (s: any) => String(s || \"\").toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n const want = norm(name);\n const movies = (m?.results || []).slice(0, 6).map((x: any) => ({ ...x, _isMovie: true }));\n const shows = (t?.results || []).slice(0, 6).map((x: any) => ({ ...x, _isMovie: false }));\n const all = [...movies, ...shows];\n const exact = all.filter(x => norm(x.title || x.name) === want || norm(x.original_title || x.original_name) === want);\n const pool = exact.length ? exact : all;\n pool.sort((a: any, b: any) => (b.popularity || 0) - (a.popularity || 0));\n return pool[0] || null;\n}\n\nasync function tmdbById(id: string, type: string) {\n const r = await fetch(SHASHTI + \"/tmdb/\" + (type === \"tv\" ? \"tv\" : \"movie\") + \"/\" + id, { signal: AbortSignal.timeout(15000) }).then(x => x.ok ? x.json() : null).catch(() => null);\n if (!r?.id) return null;\n return { ...r, _isMovie: type !== \"tv\" };\n}\n\nexport async function findByPlot(raw: string) {\n const t0 = Date.now();\n const dbg: any = {};\n const arabic = /[\\u0600-\\u06FF]/.test(raw.trim());\n let en = raw.trim();\n\n if (arabic) {\n try {\n const tr = await fetch(SHASHTI + \"/translate?q=\" + encodeURIComponent(raw.trim()) + \"&target=en\", { signal: AbortSignal.timeout(15000) })\n .then(r => r.ok ? r.json() : null);\n if (tr?.text) en = String(tr.text);\n } catch { /* نكمل بالنص كما هو */ }\n }\n dbg.translated = en;\n\n /* ═══ (أ) ثلاثة بحوث متوازية — صياغات مختلفة، إجماعها أوثق من ترتيب بحث واحد ═══\n ١) الإنجليزي + \"movie or tv series\" (الصياغة الأصلية، أثبتت نجاحها)\n ٢) الإنجليزي عارياً + \"what movie is\" (صياغة السؤال، تلتقط منتديات وأسئلة الناس)\n ٣) العربي كما كتبه (مقالات عربية تصف الأعمال بلغة المستخدم نفسها) */\n const queries: { tag: string; q: string; w: number }[] = [\n { tag: \"en\", q: en + \" movie or tv series\", w: 1.0 },\n { tag: \"ask\", q: \"what movie or series is this: \" + en, w: 0.9 },\n ];\n if (arabic) queries.push({ tag: \"ar\", q: raw.trim() + \" فلم او مسلسل\", w: 0.8 });\n else queries.push({ tag: \"plot\", q: en + \" plot summary film\", w: 0.8 });\n dbg.queries = queries.map(x => x.q);\n\n const runs = await Promise.all(queries.map(async x => {\n try { return { tag: x.tag, w: x.w, results: await searx(x.q) }; }\n catch (e) { return { tag: x.tag, w: x.w, results: [] as any[], err: String(e).slice(0, 120) }; }\n }));\n dbg.perQuery = runs.map(r => ({ tag: r.tag, n: r.results.length, err: (r as any).err }));\n\n const map = new Map();\n let allText = \"\";\n for (const r of runs) {\n harvest(map, r.results, r.tag, r.w);\n allText += \" \" + r.results.slice(0, 20).map((x: any) => (x?.title || \"\") + \" \" + (x?.content || \"\")).join(\" \");\n }\n\n let cands = [...map.values()];\n countMentions(cands, allText);\n\n /* الدمج النهائي للوزن:\n - وزن الموقع (score) الأساس\n - ×(١ + عدد البحوث اللي ظهر فيها) ← تصويت (أ)\n - + جذر عدد التكرارات ← (ج) */\n for (const c of cands) {\n const vote = 1 + (c.queries.size - 1) * 1.2;\n (c as any).final = c.score * vote + Math.sqrt(c.mentions) * 0.35;\n }\n cands.sort((a: any, b: any) => b.final - a.final);\n cands = cands.slice(0, 8);\n dbg.candidates = cands.map((c: any) => ({\n title: c.title, final: +c.final.toFixed(2), pos: +c.score.toFixed(2),\n votes: [...c.queries].join(\"+\"), mentions: c.mentions,\n }));\n\n /* ═══ تحويل لبطاقات + (ب) التحقّق من الملخص ═══ */\n const qWords = new Set(words(en));\n const items: any[] = [];\n const seen = new Set();\n const rejected: any[] = [];\n\n /* الحل بالتوازي لا بالتسلسل — ثمانية مرشَّحين تسلسلياً كانوا ~١٣ث وحدهم (مقاس) */\n const resolved = await Promise.all(cands.map(async (c) => {\n try {\n if (c.imdb) return { c, hit: await tmdbFind(c.imdb) };\n if (c.tmdb) return { c, hit: await tmdbById(c.tmdb, c.tmdbType || \"movie\") };\n if (c.title) return { c, hit: await tmdbSearch(c.title) };\n } catch { /* مرشَّح فشل، نكمل */ }\n return { c, hit: null as any };\n }));\n\n for (const { c, hit } of resolved) {\n if (items.length >= 5) break;\n if (!hit?.id || seen.has(hit.id)) continue;\n seen.add(hit.id);\n\n /* (ب) التحقّق: كم كلمة معنى مشتركة بين وصفك وملخّص العمل؟ */\n const ovWords = new Set(words((hit.overview || \"\") + \" \" + (hit.title || hit.name || \"\")));\n let shared = 0;\n for (const w of qWords) if (ovWords.has(w)) shared++;\n const overlap = qWords.size ? shared / qWords.size : 0;\n\n const item = {\n id: hit.id,\n imdb: c.imdb || null,\n isMovie: !!hit._isMovie,\n title: hit.title || hit.name || c.title,\n year: String(hit.release_date || hit.first_air_date || \"\").slice(0, 4) || null,\n poster: hit.poster_path ? \"https://image.tmdb.org/t/p/w185\" + hit.poster_path : null,\n overview: hit.overview || \"\",\n rating: hit.vote_average ? Math.round(hit.vote_average * 10) / 10 : null,\n votes: [...c.queries].join(\"+\"),\n voteCount: c.queries.size,\n mentions: c.mentions,\n shared,\n overlap: +overlap.toFixed(2),\n base: +(c as any).final.toFixed(2),\n final: 0,\n why: c.why,\n };\n\n /* الملخّص بصفر كلمة مشتركة **وبلا إجماع** = مرشَّح ضعيف يُسقَط.\n (لو ظهر ببحثين أو أكثر نبقيه — الإجماع يغطّي ملخصاً فقيراً بالكلمات.) */\n if (shared === 0 && c.queries.size < 2 && c.mentions < 3) {\n rejected.push({ title: item.title, reason: \"صفر تطابق بالملخص وبلا إجماع\" });\n continue;\n }\n\n item.final = +((c as any).final * (1 + overlap * 1.5)).toFixed(2);\n items.push(item);\n }\n\n /* الترتيب النهائي بعد التحقّق — التطابق بالملخص يقدر يقلب الترتيب */\n items.sort((a, b) => b.final - a.final);\n dbg.rejected = rejected;\n dbg.ms = Date.now() - t0;\n return { items: items.slice(0, 4), debug: dbg };\n}\n",[],[690],"afe77f16-63be-4d07-87fe-88e7ff065bdf","embtest.ts","07940258-aff8-11f1-b3bc-1607ee4eb77e",70,["D",1789361552552],"/* محاولة ثانية: نسخة أقدم من المكتبة (onnxruntime-web) — بعض البيئات تدعمها وحدها */\nexport default async function (req: Request): Promise {\n const out: any = {};\n const variants = [\n \"https://esm.sh/@xenova/transformers@2.17.2\",\n \"https://esm.sh/@huggingface/transformers@3.7.5\",\n ];\n for (const url of variants) {\n const o: any = {};\n const t0 = Date.now();\n try {\n const mod: any = await import(url);\n o.import = Date.now() - t0;\n if (mod.env) { mod.env.allowLocalModels = false; mod.env.useBrowserCache = false; }\n const t1 = Date.now();\n const ex: any = await mod.pipeline(\"feature-extraction\", \"Xenova/all-MiniLM-L6-v2\");\n o.load = Date.now() - t1;\n const t2 = Date.now();\n const e = await ex([\"a family lives in silence\"], { pooling: \"mean\", normalize: true });\n o.embed = Date.now() - t2;\n o.dims = e.dims;\n o.ok = true;\n } catch (e) { o.err = String(e).slice(0, 200); }\n out[url] = o;\n if (o.ok) break;\n }\n return Response.json(out);\n}\n",[],[697],"70999392-36a8-46b7-9281-0b01df3cbbfc","diag.ts","8ef0ba88-b071-11f1-bc59-1607ee4eb77e",202,["D",1789498659705],"/* لماذا لم يتغيّر شيء؟ هل النقد موجود لهذه الأفلام أصلاً؟ */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\n\nexport default async function (req: Request): Promise {\n const out: any = {};\n for (const t of [\"Joker\", \"1917\", \"The Matrix\", \"Whiplash\"]) {\n const r = await sqlite.execute({\n sql: `SELECT title, year, length(crit), length(plots) FROM plotC WHERE title = ? LIMIT 2`,\n args: [t],\n });\n out[t] = (r.rows || []).length\n ? (r.rows || []).map((x: any) => ({ سنة: x[1], \"طول النقد\": x[2], \"طول الملخّصات\": x[3] }))\n : \"❌ غير موجود بالفهرس (خارج أول ٢٤٠٠)\";\n }\n /* كم عملاً بالفهرس لديه نقد فعلاً؟ */\n const c = await sqlite.execute(`SELECT count(*) FROM plotC WHERE length(crit) > 100`);\n out[\"أعمال لديها نقد بالفهرس\"] = c.rows[0][0];\n /* هل كلمة stair موجودة بنقد أي فلم؟ */\n const s = await sqlite.execute(`SELECT title FROM plotC WHERE crit LIKE '%stair%' LIMIT 5`);\n out[\"نقود تذكر stair\"] = (s.rows || []).map((x: any) => x[0]);\n return Response.json(out);\n}\n",[],[705],"48217d0c-ba7e-4fd8-bb7a-148cd53d1a6b","deep.ts","03de0e02-b01a-11f1-9a6d-1607ee4eb77e",165,["D",1789484646618],"/* توسيع قائمة المرشّحين: هل قراءة ٢٥٠ بدل ٤٠ تُدخل الحالات المستعصية؟\n المنطق المقاس: Forrest Gump بالمرتبة ١٠٨ لفظياً — خارج الأربعين، وداخل المئتين. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst KEY = () => Deno.env.get(\"JINA_KEY2\") || \"\";\n\nasync function rerank(query: string, docs: string[], top = 20) {\n for (let a = 0; a < 2; a++) {\n try {\n const r = await fetch(\"https://api.jina.ai/v1/rerank\", {\n method: \"POST\",\n headers: { \"content-type\": \"application/json\", authorization: \"Bearer \" + KEY() },\n body: JSON.stringify({ model: \"jina-reranker-v2-base-multilingual\", query, documents: docs, top_n: top }),\n signal: AbortSignal.timeout(45000),\n });\n const j = await r.json();\n if (j?.results) return { order: j.results.map((x: any) => x.index as number), tokens: j?.usage?.total_tokens || 0 };\n } catch { /* نعيد */ }\n await new Promise(s => setTimeout(s, 3000));\n }\n return null;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"6\");\n const sizes = (u.searchParams.get(\"sizes\") || \"40,250\").split(\",\").map(Number);\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"لفظي\"].concat(sizes.map(s => \"ترتيب \" + s));\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n let tokens = 0, errs = 0;\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const max = Math.max(...sizes);\n\n const r = await sqlite.execute({\n sql: `SELECT title, year, kind, substr(plots,1,380), substr(kw,1,200)\n FROM plot5 WHERE plot5 MATCH ? ORDER BY bm25(plot5,1,1,1) LIMIT ${max}`, args: [match] });\n const pool = (r.rows || []).map((x: any) => ({\n title: String(x[0]),\n doc: `${x[0]} (${x[1]}, ${x[2] === \"tv\" ? \"TV series\" : \"film\"}). ${String(x[3] || \"\")} Keywords: ${String(x[4] || \"\")}`,\n }));\n\n const want = norm(exp);\n const rk = (l: string[]) => l.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n const lists: Record = { \"لفظي\": pool.map(c => c.title).slice(0, 20) };\n const deepPos = rk(pool.map(c => c.title));\n\n for (const s of sizes) {\n const sub = pool.slice(0, s);\n if (sub.length < 2) { lists[\"ترتيب \" + s] = sub.map(c => c.title); continue; }\n const rr = await rerank(ar, sub.map(c => c.doc), 20);\n if (rr) { tokens += rr.tokens; lists[\"ترتيب \" + s] = rr.order.map((i: number) => sub[i].title); }\n else { errs++; lists[\"ترتيب \" + s] = sub.map(c => c.title).slice(0, 20); }\n await new Promise(x => setTimeout(x, 1200));\n }\n\n const parts: string[] = [];\n for (const v of V) {\n const k = rk(lists[v]);\n if (k === 0) tally[v][0]++;\n if (k >= 0 && k < 5) tally[v][1]++;\n if (k >= 0 && k < 20) tally[v][2]++;\n parts.push(k < 0 ? \"خارج\" : String(k + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")} [لفظياً: ${deepPos < 0 ? \"خارج \" + max : \"#\" + (deepPos + 1)}]`);\n }\n\n const n = rows.length;\n return Response.json({\n رموز: tokens, أخطاء: errs, \"لكل بحثة\": Math.round(tokens / (n || 1)),\n أعمدة: V.join(\" · \"),\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[713],"6d53cdc0-37bd-45b4-a642-16f7811dca49","critpull.ts","f3110098-b136-11f1-8fab-1607ee4eb77e",199,["D",1789498498936],"/* سحب نقد إيبرت لعيّنة قياس — ٣٠٠ فلم تشمل حالات الاختبار */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst UA = { \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/120 Safari/537.36\" };\n\nfunction clean(html: string) {\n return html.replace(//gi, \" \").replace(//gi, \" \")\n .replace(/<[^>]+>/g, \" \").replace(/&[a-z#0-9]+;/g, \" \").replace(/\\s+/g, \" \").trim();\n}\nconst slug = (s: string) => String(s || \"\").toLowerCase().replace(/[^a-z0-9\\s-]/g, \"\").trim().replace(/\\s+/g, \"-\");\n\nasync function ebert(title: string, year: string) {\n const tries = [\n `https://www.rogerebert.com/reviews/${slug(title)}-movie-review-${year}`,\n `https://www.rogerebert.com/reviews/${slug(title)}-${year}`,\n `https://www.rogerebert.com/reviews/${slug(title)}`,\n ];\n for (const u of tries) {\n for (let a = 0; a < 2; a++) {\n try {\n const r = await fetch(u, { headers: UA, signal: AbortSignal.timeout(11000) });\n if (r.status === 429 || r.status >= 500) { await new Promise(s => setTimeout(s, 1200)); continue; }\n if (!r.ok) break;\n const t = clean(await r.text());\n /* نقتطع الترويسة والتذييل: النقد الفعلي يبدأ بعد اسم الكاتب */\n if (t.length > 2000) return t.slice(0, 9000);\n break;\n } catch { await new Promise(s => setTimeout(s, 600)); }\n }\n }\n return \"\";\n}\n\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; await fn(items[k]); }\n }));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const n = Number(u.searchParams.get(\"n\") || \"120\");\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const norm = (s: string) => String(s || \"\").toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n /* أولوية: أعمال الاختبار أولاً، ثم أفلام شائعة — لضمان قياس ذي معنى */\n const wanted = new Set(CASES.map(c => norm(c[1])));\n const priority = all.filter(r => r.kind === \"movie\" && wanted.has(norm(r.title)) && r.crit === undefined);\n const rest = all.filter(r => r.kind === \"movie\" && !wanted.has(norm(r.title)) && r.crit === undefined);\n const todo = [...priority, ...rest].slice(0, n);\n\n const t0 = Date.now();\n let ok = 0, chars = 0;\n await pool(todo, 5, async (r: any) => {\n const t = await ebert(r.title, r.year);\n r.crit = t;\n if (t) { ok++; chars += t.length; }\n await new Promise(s => setTimeout(s, 250));\n });\n\n await blob.setJSON(\"corpus_v2\", all);\n const done = all.filter(r => r.crit !== undefined).length;\n const withCrit = all.filter(r => r.crit).length;\n return Response.json({\n عولج: todo.length, نجح: ok, \"متوسط\": ok ? Math.round(chars / ok) + \" حرف\" : \"—\",\n ms: Date.now() - t0, \"أُنجز\": done, \"لديه نقد\": withCrit,\n });\n}\n",[],[721],"03f92a58-d7ba-46a8-87b5-b1393cad7e00","critics.ts","bb67a2be-b136-11f1-aad7-1607ee4eb77e",198,["D",1789498405558],"/* قياس جدوى النقد السينمائي كمصدر: التغطية أولاً — قبل أي بناء */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst UA = { \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/120 Safari/537.36\" };\n\nfunction clean(html: string) {\n return html.replace(//gi, \" \").replace(//gi, \" \")\n .replace(/<[^>]+>/g, \" \").replace(/&[a-z#0-9]+;/g, \" \").replace(/\\s+/g, \" \").trim();\n}\nconst slug = (s: string) => String(s || \"\").toLowerCase()\n .replace(/[^a-z0-9\\s-]/g, \"\").trim().replace(/\\s+/g, \"-\");\n\n/* إيبرت: نجرّب صيغتي الرابط الشائعتين */\nasync function ebert(title: string, year: string) {\n const tries = [\n `https://www.rogerebert.com/reviews/${slug(title)}-movie-review-${year}`,\n `https://www.rogerebert.com/reviews/${slug(title)}-${year}`,\n `https://www.rogerebert.com/reviews/${slug(title)}`,\n ];\n for (const u of tries) {\n try {\n const r = await fetch(u, { headers: UA, signal: AbortSignal.timeout(12000) });\n if (!r.ok) continue;\n const t = clean(await r.text());\n if (t.length > 2000) return { url: u, len: t.length, text: t };\n } catch { /* نكمل */ }\n }\n return null;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const n = Number(u.searchParams.get(\"n\") || \"12\");\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n\n const C: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n /* عيّنة موزّعة على الأفلام فقط (إيبرت لا يغطّي المسلسلات) */\n const movies = C.filter(r => r.kind === \"movie\");\n const step = Math.max(1, Math.floor(movies.length / 60));\n const sample = Array.from({ length: n }, (_, i) => movies[(from + i) * step]).filter(Boolean);\n\n const rows: string[] = [];\n let ok = 0, chars = 0;\n for (const r of sample) {\n const e = await ebert(r.title, r.year);\n if (e) { ok++; chars += e.len; }\n rows.push(`${e ? \"✅\" : \"❌\"} ${r.title} (${r.year})${e ? \" — \" + e.len + \" حرف\" : \"\"}`);\n }\n\n return Response.json({\n \"عيّنة أفلام\": sample.length,\n \"التغطية\": ok + \"/\" + sample.length + \" (\" + Math.round(ok / sample.length * 100) + \"٪)\",\n \"متوسط طول النقد\": ok ? Math.round(chars / ok) + \" حرف\" : \"—\",\n التفصيل: rows,\n });\n}\n",[],[729],"84bc654f-c6cc-486c-beec-e755a30acf0f","critbench.ts","495c5f6a-b137-11f1-b7c2-1607ee4eb77e",211,["D",1789499063845],"/* القياس: عمود النقد — هل يحلّ «رقص على الدرج» بلا إضرار بالحالات العادية؟ */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\n/* حالات المشاهد — الفشل الذي نحاول حلّه */\nconst SCENE: [string, string][] = [\n [\"a movie with a scene of a man dancing on a long colorful staircase\", \"Joker\"],\n [\"a movie shot in one long continuous take without cuts\", \"1917\"],\n [\"a movie where a drummer bleeds and his teacher throws a chair at him\", \"Whiplash\"],\n [\"a movie with a bullet time scene where he dodges bullets bending backwards\", \"The Matrix\"],\n];\n\nconst W: [string, string][] = [\n [\"بلا وصف سمعي\", \"bm25(plotC,1,1,1,2,2,0)\"],\n [\"وصف ٠٫٣\", \"bm25(plotC,1,1,1,2,2,0.3)\"],\n [\"وصف ٠٫٧\", \"bm25(plotC,1,1,1,2,2,0.7)\"],\n [\"وصف ١٫٥\", \"bm25(plotC,1,1,1,2,2,1.5)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const mode = u.searchParams.get(\"mode\") || \"scene\";\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const scope = await sqlite.execute(`SELECT title FROM plotC`);\n const allowed = new Set((scope.rows || []).map((x: any) => String(x[0])));\n\n const list: [string, string][] = mode === \"scene\" ? SCENE : (CASES as any);\n const tally: any = {}; for (const [n] of W) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n let tested = 0;\n\n for (const [ar, exp] of list) {\n const want = norm(exp);\n if (![...allowed].some(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); })) continue;\n if (tested >= 22) break;\n tested++;\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const parts: string[] = [];\n\n for (const [name, rank] of W) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM plotC WHERE plotC MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n const l = (r.rows || []).map((x: any) => String(x[0]));\n const rk = l.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n return Response.json({\n الوضع: mode === \"scene\" ? \"حالات المشاهد\" : \"الحالات العادية\",\n \"نطاق المقارنة\": allowed.size, مختبَر: tested,\n أعمدة: W.map(w => w[0]).join(\" · \"),\n النتيجة: Object.fromEntries(W.map(([n]) =>\n [n, `أول ${tally[n][0]}/${tested} · ضمن٥ ${tally[n][1]}/${tested} · ضمن٢٠ ${tally[n][2]}/${tested}`])),\n التفصيل: rows,\n });\n}\n",[],[737],"7d744df7-1144-428d-96f7-18b9d3154605","coverage.ts","71fcfd6e-b072-11f1-8f30-1607ee4eb77e",87,["D",1789414100964],"/* فحص تغطية القاعدة الجديدة على أعمال لم تُضَف يدوياً */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst CHECK = [\"Home Alone\", \"Bicycle Thieves\", \"The Godfather\", \"Casablanca\", \"Jaws\",\n \"Rocky\", \"Back to the Future\", \"E.T.\", \"The Shining\", \"Speed\", \"Das Boot\",\n \"Ratatouille\", \"Toy Story\", \"Kill Bill\", \"Gladiator\", \"Life of Pi\", \"The Martian\",\n \"Groundhog Day\", \"The Sixth Sense\", \"Up\", \"Titanic\", \"The Matrix\", \"Memento\",\n \"Forrest Gump\", \"Friends\", \"Seinfeld\", \"Lost\", \"The Office\", \"Twin Peaks\"];\n\nexport default async function (req: Request): Promise {\n const v2: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const v1: any[] = (await blob.getJSON(\"imdbcorpus_v1\")) as any[] || [];\n const norm = (s: string) => String(s || \"\").toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n const inSet = (C: any[], name: string) => {\n const w = norm(name);\n const hits = C.filter(r => { const t = norm(r.title); return t === w || t.startsWith(w + \" \") || t === w; });\n return hits.length ? hits.map(h => `${h.title} (${h.year})`).join(\" · \") : null;\n };\n\n const rows = CHECK.map(n => ({\n name: n,\n \"القديمة\": inSet(v1, n) ? \"✅\" : \"❌\",\n \"الجديدة\": inSet(v2, n) ? \"✅ \" + inSet(v2, n)!.slice(0, 40) : \"❌\",\n }));\n\n const years = v2.map(r => Number(r.year) || 0).filter(y => y > 1900);\n return Response.json({\n \"قاعدة جديدة\": v2.length,\n \"قبل ٢٠٠٠\": years.filter(y => y < 2000).length,\n \"قبل ١٩٩٠\": years.filter(y => y < 1990).length,\n \"تغطية القديمة\": rows.filter(r => r[\"القديمة\"] === \"✅\").length + \"/\" + CHECK.length,\n \"تغطية الجديدة\": rows.filter(r => r[\"الجديدة\"].startsWith(\"✅\")).length + \"/\" + CHECK.length,\n \"المفقود بالجديدة\": rows.filter(r => r[\"الجديدة\"] === \"❌\").map(r => r.name),\n \"تفصيل\": rows,\n });\n}\n",[],[745],"2f427439-081b-4e0e-8b33-430842bd3fb8","cov.ts","bc955566-b080-11f1-8581-1607ee4eb77e",107,["D",1789420239132],"/* الإصلاح الجذري: الترتيب بتغطية الكلمات المختلفة أولاً، ثم BM25.\n السبب المقاس: BM25 مع OR يكافئ تكرار كلمة واحدة لا تغطية كلمات مختلفة —\n فسقط Back to the Future رغم احتوائه ٨٨٪ من كلمات الاستعلام. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst esc = (w: string) => w.replace(/'/g, \"''\");\n\nasync function baseline(match: string) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM plot5 WHERE plot5 MATCH ? ORDER BY bm25(plot5,1,1,1) LIMIT 20`,\n args: [match],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\n/* التغطية تُحسب داخل قاعدة البيانات نفسها — بلا نقل نصوص للخارج */\nasync function coverage(match: string, terms: string[], mode: \"strict\" | \"blend\") {\n const cov = terms.map(w =>\n `(CASE WHEN instr(lower(plots||' '||kw||' '||title), '${esc(w)}') > 0 THEN 1 ELSE 0 END)`).join(\" + \");\n const order = mode === \"strict\"\n ? `cov DESC, s ASC`\n /* مزيج: التغطية إشارة قوية لكن ليست مطلقة — تمنع صعود مستند يطابق كلمات تافهة */\n : `(cov * 1.0 / ${terms.length}) * 12 + (-s) DESC`;\n const r = await sqlite.execute({\n sql: `SELECT title, (${cov}) AS cov, bm25(plot5,1,1,1) AS s\n FROM plot5 WHERE plot5 MATCH ?\n ORDER BY ${mode === \"strict\" ? \"cov DESC, s ASC\" : \"((cov * 1.0 / \" + terms.length + \") * 12 - s) DESC\"}\n LIMIT 20`,\n args: [match],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"12\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"أساس BM25\", \"تغطية صارمة\", \"تغطية ممزوجة\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n let ms = 0;\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const want = norm(exp);\n\n const t0 = Date.now();\n const lists: Record = {\n \"أساس BM25\": await baseline(match),\n \"تغطية صارمة\": await coverage(match, terms as string[], \"strict\"),\n \"تغطية ممزوجة\": await coverage(match, terms as string[], \"blend\"),\n };\n ms += Date.now() - t0;\n\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.join(\" · \"),\n \"زمن كل بحث\": Math.round(ms / (n * 3)) + \"ms\",\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[753],"d98272bf-c5c3-4224-b315-5260340fee7d","corpus.ts","dfc53fe6-affa-11f1-98ef-1607ee4eb77e",72,["D",1789362745600],"/* يضمن وجود أعمال الاختبار بالقاعدة + يخدم القاعدة للجوال مع CORS */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst KEY = \"corpus_v1\";\n\nconst WANT: [string, string][] = [\n [\"movie\", \"A Quiet Place\"], [\"tv\", \"Dark\"], [\"movie\", \"Inception\"], [\"tv\", \"Squid Game\"],\n [\"tv\", \"Severance\"], [\"movie\", \"The Sixth Sense\"], [\"movie\", \"Up\"], [\"movie\", \"Groundhog Day\"],\n [\"tv\", \"Breaking Bad\"], [\"movie\", \"The Lincoln Lawyer\"], [\"movie\", \"The Matrix\"],\n [\"movie\", \"Interstellar\"], [\"movie\", \"Parasite\"], [\"tv\", \"Chernobyl\"], [\"movie\", \"Titanic\"],\n];\n\nasync function j(u: string) {\n try { const r = await fetch(u, { signal: AbortSignal.timeout(15000) }); return r.ok ? await r.json() : null; }\n catch { return null; }\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n let all: any[] = (await blob.getJSON(KEY)) as any[] || [];\n\n if (u.searchParams.get(\"ensure\")) {\n const have = new Set(all.map(r => String(r.title || \"\").toLowerCase()));\n const added: string[] = [];\n for (const [kind, name] of WANT) {\n if (have.has(name.toLowerCase())) continue;\n const s = await j(`${SH}/tmdb/search/${kind}?query=${encodeURIComponent(name)}`);\n const x = s?.results?.[0];\n if (!x) continue;\n const d = await j(`${SH}/tmdb/${kind}/${x.id}?append_to_response=keywords`);\n const kwo = d?.keywords || {};\n all.push({\n id: x.id, kind, title: x.title || x.name,\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n kw: ((kwo.keywords || kwo.results || []) as any[]).map(w => w.name).slice(0, 20),\n genres: (d?.genres || []).map((g: any) => g.name),\n });\n added.push(name);\n }\n await blob.setJSON(KEY, all);\n return Response.json({ added, total: all.length });\n }\n\n /* نخدم نصاً مضغوطاً: سطر لكل عمل — أخف بكثير من JSON كامل للجوال */\n const lines = all.map(r => [\n r.kind, r.id, r.title, r.year,\n [r.title, (r.genres || []).join(\", \"), (r.kw || []).join(\", \"), r.ov].filter(Boolean).join(\" . \"),\n ].join(\"\\t\"));\n return new Response(lines.join(\"\\n\"), {\n headers: {\n \"content-type\": \"text/plain; charset=utf-8\",\n \"access-control-allow-origin\": \"*\",\n \"cache-control\": \"public, max-age=3600\",\n },\n });\n}\n",[],[761],"fcff2c88-89eb-4c2c-889a-24594b7cb394","compare.ts","910190ee-b072-11f1-aba4-1607ee4eb77e",90,["D",1789414291825],"/* هل التوسيع اللفظي يحل المشكلة فعلاً؟ اختبار مباشر قبل بناء أي نظام مرادفات */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\n\nasync function s(terms: string[]) {\n const r = await sqlite.execute({\n sql: `SELECT title, year, bm25(plot2,1.0,1.0) FROM plot2 WHERE plot2 MATCH ? ORDER BY 3 LIMIT 6`,\n args: [terms.map(w => `\"${w}\"`).join(\" OR \")],\n });\n return (r.rows || []).map((x: any, i: number) => `${i + 1}. ${x[0]} (${x[1]})`);\n}\nconst tok = (q: string) => (q.toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => w.length > 1);\n\nexport default async function (req: Request): Promise {\n const base = \"A child whose family travels and leaves him at home alone faces thieves\";\n const out: any = {};\n out[\"الأصل\"] = await s(tok(base));\n out[\"+ burglars\"] = await s(tok(base).concat([\"burglars\", \"burglar\"]));\n out[\"+ burglars + بلا الشائع\"] = await s(\n tok(base).filter(w => ![\"a\", \"and\", \"him\", \"at\", \"whose\", \"faces\", \"child\"].includes(w)).concat([\"burglars\", \"burglar\"]));\n\n /* اختبار ثانٍ: Matrix — حبة حمراء */\n const m = \"a movie about a virtual world and the hero takes a red pill\";\n out[\"ماتريكس الأصل\"] = await s(tok(m));\n out[\"ماتريكس + red pill معاً\"] = await s(tok(m).concat([\"matrix\", \"simulated\", \"reality\"]));\n\n return Response.json(out);\n}\n",[],[769],"08e9cfc8-9cef-46ea-89ac-738d2d5eb742","cmumerge.ts","2b18fd02-b0c3-11f1-91f2-1607ee4eb77e",148,["D",1789448771310],"/* دمج ملخّصات CMU بقاعدتنا — مطابقة بالاسم والسنة مع تسامح سنة واحدة */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst norm = (s: string) => String(s || \"\").toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const idx: Record = (await blob.getJSON(\"cmu_meta\")) as any || {};\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const out: any = { \"ملخّصات CMU\": Object.keys(idx).length, \"أعمال القاعدة\": all.length };\n\n if (u.searchParams.get(\"dry\")) {\n /* عدّ المطابقات بلا كتابة */\n let hit = 0, hitTv = 0, hitMv = 0;\n for (const r of all) {\n const t = norm(r.title), y = Number(r.year) || 0;\n let found = false;\n for (const dy of [0, -1, 1]) if (idx[t + \"|\" + (y + dy)]) { found = true; break; }\n if (found) { hit++; r.kind === \"tv\" ? hitTv++ : hitMv++; }\n }\n out[\"مطابق\"] = hit + \" (\" + Math.round(hit / all.length * 100) + \"٪)\";\n out[\"أفلام\"] = hitMv; out[\"مسلسلات\"] = hitTv;\n return Response.json(out);\n }\n\n const start = Number(u.searchParams.get(\"start\") || \"0\");\n const count = Number(u.searchParams.get(\"n\") || \"20563\");\n let added = 0, chars = 0;\n for (let i = start; i < Math.min(start + count, all.length); i++) {\n const r = all[i];\n const t = norm(r.title), y = Number(r.year) || 0;\n for (const dy of [0, -1, 1]) {\n const v = idx[t + \"|\" + (y + dy)];\n if (v) { r.cmu = v; added++; chars += v.length; break; }\n }\n }\n await blob.setJSON(\"corpus_v2\", all);\n const withCmu = all.filter(r => r.cmu).length;\n out[\"أُضيف هذه الجولة\"] = added;\n out[\"متوسط الطول\"] = added ? Math.round(chars / added) + \" حرف\" : \"—\";\n out[\"إجمالي من لديه حبكة CMU\"] = withCmu + \" (\" + Math.round(withCmu / all.length * 100) + \"٪)\";\n return Response.json(out);\n}\n",[],[777],"7e5c28d9-138d-4e77-a7c0-b28d7eb369f7","cmu.ts","eb26adfc-b0c2-11f1-a95c-1607ee4eb77e",147,["D",1789448738703],"/* تنزيل مجموعة CMU (٤٢٣٠٦ ملخّص حبكة من ويكيبيديا) وفكّ ضغطها وتخزينها.\n بديل واحد عن ٢٠ ألف نداء مخنوق لويكيبيديا. */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst URL_TGZ = \"http://www.cs.cmu.edu/~ark/personas/data/MovieSummaries.tar.gz\";\n\n/* قارئ tar بسيط: كل عنصر رأس ٥١٢ بايت ثم المحتوى مُحاذى لـ٥١٢ */\nfunction* tarEntries(buf: Uint8Array) {\n let off = 0;\n const dec = new TextDecoder();\n while (off + 512 <= buf.length) {\n const header = buf.subarray(off, off + 512);\n const name = dec.decode(header.subarray(0, 100)).replace(/\\0.*$/, \"\").trim();\n if (!name) { off += 512; continue; }\n const sizeStr = dec.decode(header.subarray(124, 136)).replace(/\\0.*$/, \"\").trim();\n const size = parseInt(sizeStr, 8) || 0;\n const start = off + 512;\n yield { name, data: buf.subarray(start, start + size) };\n off = start + Math.ceil(size / 512) * 512;\n }\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const out: any = {};\n\n if (u.searchParams.get(\"stage\") === \"meta\") {\n const m: any = await blob.getJSON(\"cmu_meta\");\n return Response.json({ \"أفلام بالفهرس\": m ? Object.keys(m).length : 0, عينة: m ? Object.entries(m).slice(0, 3) : [] });\n }\n\n const t0 = Date.now();\n const r = await fetch(URL_TGZ, { signal: AbortSignal.timeout(50000) });\n out.status = r.status;\n const gz = new Uint8Array(await r.arrayBuffer());\n out[\"حجم مضغوط\"] = gz.length;\n\n /* فكّ gzip عبر واجهة الضغط المدمجة بالمنصّة */\n const ds = new DecompressionStream(\"gzip\");\n const stream = new Blob([gz]).stream().pipeThrough(ds);\n const tar = new Uint8Array(await new Response(stream).arrayBuffer());\n out[\"حجم مفكوك\"] = tar.length;\n\n const dec = new TextDecoder();\n const files: Record = {};\n let plotsRaw = \"\", metaRaw = \"\";\n for (const e of tarEntries(tar)) {\n files[e.name] = e.data.length;\n if (/plot_summaries\\.txt$/.test(e.name)) plotsRaw = dec.decode(e.data);\n if (/movie\\.metadata\\.tsv$/.test(e.name)) metaRaw = dec.decode(e.data);\n }\n out.ملفات = files;\n out.ms = Date.now() - t0;\n\n /* الفهرس: اسم الفلم + السنة → الملخّص */\n const idById: Record = {};\n for (const line of metaRaw.split(\"\\n\")) {\n const c = line.split(\"\\t\");\n if (c.length < 4) continue;\n idById[c[0]] = { t: c[2], y: String(c[3] || \"\").slice(0, 4) };\n }\n const index: Record = {};\n let n = 0;\n for (const line of plotsRaw.split(\"\\n\")) {\n const tab = line.indexOf(\"\\t\");\n if (tab < 0) continue;\n const id = line.slice(0, tab), text = line.slice(tab + 1);\n const meta = idById[id];\n if (!meta?.t) continue;\n const key = meta.t.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim() + \"|\" + meta.y;\n if (!index[key]) { index[key] = text.slice(0, 6000); n++; }\n }\n out[\"ملخّصات مفهرسة\"] = n;\n await blob.setJSON(\"cmu_meta\", index);\n return Response.json(out);\n}\n",[],[785],"14dc5897-a27b-4bd5-a06a-1c55053ccfad","clean.ts","61333c8c-b081-11f1-8c8c-1607ee4eb77e",112,["D",1789420515302],"/* القياس: فهرس عادي (plot5) مقابل فهرس منظَّف عند الفهرسة (plot6) */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nasync function q(table: string, terms: string[]) {\n if (!terms.length) return [];\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY bm25(${table},1,1,1) LIMIT 20`,\n args: [terms.map(w => `\"${w}\"`).join(\" OR \")],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"15\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"عادي (plot5)\", \"منظَّف (plot6)\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n let ms = 0;\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const t0 = Date.now();\n const lists: Record = {\n \"عادي (plot5)\": await q(\"plot5\", terms),\n \"منظَّف (plot6)\": await q(\"plot6\", terms),\n };\n ms += Date.now() - t0;\n const want = norm(exp);\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" → \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n زمن: Math.round(ms / (n * 2)) + \"ms\",\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[793],"398d7fe3-de35-4065-97d5-fec5b6a739f1","cases.ts","9a5071f0-b019-11f1-a1ec-1607ee4eb77e",81,["D",1789375943458],"export const CASES: [string, string][] = [\n [\"عائلة تعيش بصمت كامل لأن وحوش تصطاد بالصوت\", \"A Quiet Place\"],\n [\"مسلسل ألماني عن سفر بالزمن فيه كهف وأطفال مفقودين\", \"Dark\"],\n [\"رجل يدخل بالأحلام عشان يزرع فكرة بعقل واحد\", \"Inception\"],\n [\"مسلسل كوري ناس مفلسين يلعبون ألعاب أطفال والخسران يموت\", \"Squid Game\"],\n [\"مسلسل عن شركة موظفينها يفصلون ذاكرة الشغل عن حياتهم\", \"Severance\"],\n [\"فلم عن ولد صغير يقدر يشوف الأموات\", \"The Sixth Sense\"],\n [\"فلم كرتون عن رجل عجوز يربط بيته ببالونات ويطير\", \"Up\"],\n [\"فلم عن رجل يعيش نفس اليوم مرة ورا مرة\", \"Groundhog Day\"],\n [\"مسلسل عن مدرس كيمياء يصير يصنع مخدرات\", \"Breaking Bad\"],\n [\"فلم عن محامي يشتغل من سيارته\", \"The Lincoln Lawyer\"],\n [\"فلم عن عالم افتراضي والبطل ياخذ حبة حمرا\", \"The Matrix\"],\n [\"فلم عن سفر بالفضاء عبر ثقب دودي عشان ينقذون البشرية\", \"Interstellar\"],\n [\"فلم كوري عن عائلة فقيرة تشتغل عند عائلة غنية\", \"Parasite\"],\n [\"مسلسل عن انفجار مفاعل نووي بالاتحاد السوفيتي\", \"Chernobyl\"],\n [\"فلم عن سفينة تضرب جبل جليد وتغرق\", \"Titanic\"],\n [\"مسلسل عن ملكة بريطانيا من شبابها لين كبرت\", \"The Crown\"],\n [\"فلم عن مهرج يصير مجرم بمدينة فاسدة\", \"Joker\"],\n [\"مسلسل عن أطفال يلاقون بنت بقوى خارقة ووحش من عالم ثاني\", \"Stranger Things\"],\n [\"فلم عن رجل محبوس بالسجن سنين وحفر نفق للهروب\", \"The Shawshank Redemption\"],\n [\"فلم عن ملاكم يتدرب على درج فيلادلفيا\", \"Rocky\"],\n [\"مسلسل عن عائلات تتقاتل على عرش والتنانين\", \"Game of Thrones\"],\n [\"فلم عن رجل عنده ذاكرة قصيرة يكتب على جسمه\", \"Memento\"],\n [\"فلم عن أب يدور دراجته المسروقة مع ابنه بروما\", \"Bicycle Thieves\"],\n [\"فلم عن سمكة صغيرة ضاعت وأبوها يدورها بالبحر\", \"Finding Nemo\"],\n [\"فلم عن رجل يركض بأمريكا كلها وله حياة غريبة\", \"Forrest Gump\"],\n [\"مسلسل عن ست نفر أصدقاء بشقة بنيويورك\", \"Friends\"],\n [\"مسلسل إسباني عن عصابة تسرق دار سك النقود بأقنعة\", \"Money Heist\"],\n [\"فلم رعب عن فندق وكاتب يجن بالشتاء\", \"The Shining\"],\n [\"فلم عن سيارة زمن وولد يرجع للماضي يقابل أهله\", \"Back to the Future\"],\n [\"فلم عن كائن فضائي صغير يصادق ولد ويبي يرجع بيته\", \"E.T. the Extra-Terrestrial\"],\n [\"a man is stranded alone on mars and grows potatoes\", \"The Martian\"],\n [\"a shark terrorizes a beach town and three men hunt it\", \"Jaws\"],\n [\"toys come alive when humans are not looking\", \"Toy Story\"],\n [\"a rat wants to become a chef in paris\", \"Ratatouille\"],\n [\"a woman kills bill after waking from a coma\", \"Kill Bill\"],\n [\"gladiator seeks revenge against the emperor who killed his family\", \"Gladiator\"],\n [\"a hobbit carries a ring to a volcano to destroy it\", \"The Lord of the Rings\"],\n [\"a german submarine crew during world war two under pressure\", \"Das Boot\"],\n [\"a boy and a tiger stuck on a lifeboat in the ocean\", \"Life of Pi\"],\n [\"a bus that explodes if it drives under fifty miles per hour\", \"Speed\"],\n [\"فلم عن رجل يلبس بدلة خفافيش ويحارب الجريمة\", \"The Dark Knight\"],\n [\"مسلسل عن أطباء بمستشفى وقصص حب بينهم\", \"Grey's Anatomy\"],\n [\"فلم عن حرب النجوم والسيوف الضوئية\", \"Star Wars\"],\n [\"فلم عن ديناصورات ترجع للحياة بجزيرة\", \"Jurassic Park\"],\n [\"فلم عن رجل أخضر كبير يزعل ويكسر كل شي\", \"Hulk\"],\n];\n",[],[801],"559dc4da-3c7b-455a-9389-9a49ee09e4f7","calib.ts","5ba120b0-b183-11f1-bbf4-1607ee4eb77e",224,["D",1789531316123],"/* معايرة عتبة الثقة بالبيانات: نجمع درجة الصلة لكل حالة ونقارنها بالنتيجة الفعلية */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\nconst APP = \"https://q8kamikaze--98aa950cb02c11f186041607ee4eb77e.web.val.run\";\n\n/* أوصاف وهمية: لا عمل لها بالكون — درجاتها تمثّل «لا جواب» */\nconst FAKE: string[] = [\n \"فلم عن فيل بنفسجي يعلّم المحاسبة للغواصات\",\n \"مسلسل عن حجر يتعلّم العزف على البيانو بالقطب الشمالي\",\n \"فلم عن سحابة تفتح مطعم سوشي بقاع المحيط\",\n \"مسلسل عن ملعقة تدير شركة طيران بكوكب زحل\",\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"12\");\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const store: any = (await blob.getJSON(\"calib_v1\")) || { real: [], fake: [] };\n\n if (u.searchParams.get(\"stage\") === \"report\") {\n const real = store.real as any[], fake = store.fake as any[];\n const ok1 = real.filter(x => x.rank === 0).map(x => x.score).sort((a, b) => a - b);\n const ok5 = real.filter(x => x.rank >= 0 && x.rank < 5).map(x => x.score).sort((a, b) => a - b);\n const bad = real.filter(x => x.rank < 0 || x.rank >= 5).map(x => x.score).sort((a, b) => a - b);\n const fk = fake.map((x: any) => x.score).sort((a: number, b: number) => a - b);\n const pct = (a: number[], p: number) => a.length ? a[Math.floor(a.length * p)] : null;\n\n /* أفضل عتبة: تفصل الناجح عن الوهمي بأقل خطأ */\n const cands = [0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5];\n const table = cands.map(t => ({\n عتبة: t,\n \"ناجحة تُوصف ضعيفة (خطأ)\": ok5.filter(s => s < t).length + \"/\" + ok5.length,\n \"وهمية تُوصف مقبولة (خطأ)\": fk.filter((s: number) => s >= t).length + \"/\" + fk.length,\n \"فاشلة تُوصف مقبولة\": bad.filter(s => s >= t).length + \"/\" + bad.length,\n }));\n\n return Response.json({\n \"عيّنات حقيقية\": real.length, \"عيّنات وهمية\": fake.length,\n \"المركز الأول — المدى\": ok1.length ? `${ok1[0].toFixed(2)} – ${ok1[ok1.length - 1].toFixed(2)} (وسيط ${pct(ok1, 0.5)?.toFixed(2)})` : \"—\",\n \"ضمن ٥ — أدنى ١٠٪\": pct(ok5, 0.1)?.toFixed(2),\n \"الفاشلة — المدى\": bad.length ? `${bad[0].toFixed(2)} – ${bad[bad.length - 1].toFixed(2)}` : \"—\",\n \"الوهمية — المدى\": fk.length ? `${fk[0].toFixed(2)} – ${fk[fk.length - 1].toFixed(2)}` : \"—\",\n جدول: table,\n });\n }\n\n /* جمع البيانات */\n const isFake = u.searchParams.get(\"fake\") === \"1\";\n const list: [string, string][] = isFake ? FAKE.map(f => [f, \"—\"]) as any : CASES.slice(from, to);\n\n for (const [ar, exp] of list) {\n try {\n const j = await fetch(APP + \"/api?q=\" + encodeURIComponent(ar), { signal: AbortSignal.timeout(28000) }).then(r => r.json());\n const titles = (j.items || []).map((x: any) => String(x.title || \"\"));\n const want = norm(exp);\n const rank = isFake ? -1 : titles.findIndex(t => {\n const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x);\n });\n const rec = { q: ar.slice(0, 40), exp, rank, score: j?.debug?.topScore ?? 0, conf: j?.debug?.conf };\n (isFake ? store.fake : store.real).push(rec);\n } catch { /* نكمل */ }\n }\n await blob.setJSON(\"calib_v1\", store);\n return Response.json({ \"حقيقية مجمَّعة\": store.real.length, \"وهمية مجمَّعة\": store.fake.length });\n}\n",[],[809],"bd302376-104b-4bf0-b0b3-ecfcb6237af2","build2.ts","c570fafa-b071-11f1-a003-1607ee4eb77e",116,["D",1789420815537],"/* بناء القاعدة بمعيار \"الأكثر تقييماً عبر التاريخ\" لا \"الأشهر الآن\".\n السبب المقاس: معيار الشعبية أسقط Home Alone الأصلي وأبقى أعمالاً مغمورة حديثة. */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\nconst KEY = \"corpus_v2\";\n\nasync function j(u: string) {\n try { const r = await fetch(u, { signal: AbortSignal.timeout(15000) }); return r.ok ? await r.json() : null; }\n catch { return null; }\n}\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n const out: R[] = []; let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; out[k] = await fn(items[k]); }\n }));\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"1\");\n const to = Number(u.searchParams.get(\"to\") || \"10\");\n const sort = u.searchParams.get(\"sort\") || \"vote_count.desc\";\n const minVotes = u.searchParams.get(\"mv\") || \"50\";\n\n const all: any[] = (await blob.getJSON(KEY)) as any[] || [];\n const seen = new Set(all.map((r: any) => r.kind + \":\" + r.id));\n\n const pages: [string, number][] = [];\n for (let p = from; p <= to; p++) { pages.push([\"movie\", p]); pages.push([\"tv\", p]); }\n\n const base = (await pool(pages, 14, async ([kind, p]) => {\n const d = await j(`${SH}/tmdb/discover/${kind}?sort_by=${sort}&page=${p}&vote_count.gte=${minVotes}`);\n return (d?.results || []).map((x: any) => ({\n id: x.id, kind, title: x.title || x.name,\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n votes: x.vote_count || 0, pop: x.popularity || 0,\n }));\n })).flat().filter((r: any) => !seen.has(r.kind + \":\" + r.id));\n\n const withImdb = await pool(base, 24, async (r: any) => {\n const e = await j(`${SH}/tmdb/${r.kind}/${r.id}/external_ids`);\n r.imdb = e?.imdb_id || null;\n return r;\n });\n const ok = withImdb.filter((r: any) => r.imdb);\n\n const groups: any[][] = [];\n for (let i = 0; i < ok.length; i += 5) groups.push(ok.slice(i, i + 5));\n await pool(groups, 12, async (g: any[]) => {\n const alias = g.map((r, i) => `t${i}: title(id: ${JSON.stringify(r.imdb)}) {\n plots(first: 6) { edges { node { plotText { plainText } } } } }`).join(\" \");\n const res = await gql(`query { ${alias} }`);\n g.forEach((r, i) => {\n r.plots = ((res?.data?.[`t${i}`]?.plots?.edges) || [])\n .map((e: any) => String(e.node?.plotText?.plainText || \"\")).filter(Boolean);\n });\n return g;\n });\n\n const merged = all.concat(ok);\n await blob.setJSON(KEY, merged);\n const wp = merged.filter((r: any) => (r.plots || []).length);\n return Response.json({\n added: ok.length, total: merged.length,\n withPlots: wp.length,\n قبل_٢٠٠٠: merged.filter((r: any) => Number(r.year) < 2000).length,\n });\n}\n",[],[817],"35cd0857-e581-43e0-ac6c-d0d323cf9ed9","build.ts","bf6cdb32-affa-11f1-97d5-1607ee4eb77e",71,["D",1789362691334],"/* يبني قاعدة القياس داخل Val Town (ملخص + وسوم بشرية) ويخزّنها بالبلوب.\n السبب: الجوال يحتاج ملفاً واحداً يحمّله، والبناء هنا يتجنّب نقل ٣ ميجا يدوياً. */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst KEY = \"corpus_v1\";\n\nasync function j(u: string) {\n for (let i = 0; i < 2; i++) {\n try {\n const r = await fetch(u, { signal: AbortSignal.timeout(15000) });\n if (r.ok) return await r.json();\n } catch { /* نعيد المحاولة */ }\n }\n return null;\n}\n\nasync function page(kind: string, p: number) {\n const d = await j(`${SH}/tmdb/discover/${kind}?sort_by=popularity.desc&page=${p}&vote_count.gte=100`);\n return (d?.results || []).map((x: any) => ({\n id: x.id, kind,\n title: x.title || x.name,\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n })).filter((r: any) => r.ov.length > 40);\n}\n\nasync function details(r: any) {\n const d = await j(`${SH}/tmdb/${r.kind}/${r.id}?append_to_response=keywords`);\n if (!d) return r;\n const kwo = d.keywords || {};\n r.kw = ((kwo.keywords || kwo.results || []) as any[]).map(w => w.name).slice(0, 20);\n r.genres = (d.genres || []).map((g: any) => g.name);\n return r;\n}\n\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n const out: R[] = []; let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; out[k] = await fn(items[k]); }\n }));\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"1\");\n const to = Number(u.searchParams.get(\"to\") || \"10\");\n\n let prev: any[] = [];\n try { prev = (await blob.getJSON(KEY)) as any[] || []; } catch { /* أول مرة */ }\n\n const pages: [string, number][] = [];\n for (let p = from; p <= to; p++) { pages.push([\"movie\", p]); pages.push([\"tv\", p]); }\n const got = (await pool(pages, 8, ([k, p]) => page(k, p))).flat();\n\n const seen = new Set(prev.map((r: any) => r.kind + \":\" + r.id));\n const fresh = got.filter((r: any) => !seen.has(r.kind + \":\" + r.id));\n const enriched = await pool(fresh, 14, details);\n\n const all = prev.concat(enriched);\n await blob.setJSON(KEY, all);\n return Response.json({ pagesFrom: from, pagesTo: to, added: enriched.length, total: all.length });\n}\n",[],[825],"9abd5baf-8e1f-4b11-bf99-18bacce08b22","boost.ts","2c0ae980-b0d9-11f1-a2ce-1607ee4eb77e",160,["D",1789458243523],"/* رفع النسبة على القاعدة الكاملة (٢٠٥٦٣):\n أ) ضبط أوزان الأعمدة — ضُبطت سابقاً على ٥٥٠٠ فقط، والنطاق يغيّر الأمثل\n ب) مكافأة العبارات المركّبة: \"time travel\" كعبارة أقوى دلالة من كلمتين متفرّقتين */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been \" +\n \"about after who that this its their he she it they them his her one two three \" +\n \"movie film series show season episode not no can will just has have had\").split(\" \"));\n\nasync function run(match: string, rank: string) {\n try {\n const r = await sqlite.execute({\n sql: `SELECT title FROM plot5 WHERE plot5 MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n return (r.rows || []).map((x: any) => String(x[0]));\n } catch { return []; }\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"23\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n /* الصيغ: [الاسم, أوزان(عنوان,ملخّص,وسم), هل نضيف عبارات؟] */\n const V: [string, string, boolean][] = [\n [\"الحالي ١:١:١\", \"1,1,1\", false],\n [\"بلا عنوان ٠:١:١\", \"0,1,1\", false],\n [\"ملخّص٢ وسم١\", \"1,2,1\", false],\n [\"ملخّص١ وسم٢\", \"1,1,2\", false],\n [\"عنوان٣ ١:١\", \"3,1,1\", false],\n [\"الحالي + عبارات\", \"1,1,1\", true],\n [\"عنوان٣ + عبارات\", \"3,1,1\", true],\n ];\n\n const tally: any = {}; for (const [n] of V) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const words = (en.toLowerCase().match(/[a-z0-9]+/g) || []).filter(w => w.length > 1);\n const uniq = [...new Set(words)];\n const base = uniq.map(w => `\"${w}\"`).join(\" OR \");\n\n /* العبارات: كل زوج متجاور كلمتاه ذواتا معنى — تُضاف كمطابقة عبارة */\n const phrases: string[] = [];\n for (let i = 0; i + 1 < words.length; i++) {\n if (STOP.has(words[i]) || STOP.has(words[i + 1])) continue;\n phrases.push(`\"${words[i]} ${words[i + 1]}\"`);\n }\n /* العبارة تُضاف مرتين لترجيحها داخل نفس الاستعلام */\n const withPhrases = phrases.length\n ? base + \" OR \" + phrases.join(\" OR \") + \" OR \" + phrases.join(\" OR \")\n : base;\n\n const want = norm(exp);\n const parts: string[] = [];\n for (const [name, w, usePhrase] of V) {\n const list = await run(usePhrase ? withPhrases : base, `bm25(plot5,${w})`);\n const rk = list.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.map(v => v[0]).join(\" · \"),\n النتيجة: Object.fromEntries(V.map(([name]) =>\n [name, `أول ${tally[name][0]}/${n} · ضمن٥ ${tally[name][1]}/${n} · ضمن٢٠ ${tally[name][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[833],"1258dcb6-2284-4713-92a0-b4b311ff581c","bigbench.ts","c352b898-b018-11f1-aa27-1607ee4eb77e",78,["D",1789375582772],"/* قياس موسَّع: ٤٥ عينة — عربية وإنجليزية · مشهورة وغامضة · صيغ وصف مختلفة */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\n\ntype Case = [string, string, string]; /* الوصف · العمل المتوقَّع · التصنيف */\nconst CASES: Case[] = [\n /* ═══ عربي · مشهور ═══ */\n [\"عائلة تعيش بصمت كامل لأن وحوش تصطاد بالصوت\", \"A Quiet Place\", \"عربي/مشهور\"],\n [\"مسلسل ألماني عن سفر بالزمن فيه كهف وأطفال مفقودين\", \"Dark\", \"عربي/مشهور\"],\n [\"رجل يدخل بالأحلام عشان يزرع فكرة بعقل واحد\", \"Inception\", \"عربي/مشهور\"],\n [\"مسلسل كوري ناس مفلسين يلعبون ألعاب أطفال والخسران يموت\", \"Squid Game\", \"عربي/مشهور\"],\n [\"مسلسل عن شركة موظفينها يفصلون ذاكرة الشغل عن حياتهم\", \"Severance\", \"عربي/مشهور\"],\n [\"فلم عن ولد صغير يقدر يشوف الأموات\", \"The Sixth Sense\", \"عربي/مشهور\"],\n [\"فلم كرتون عن رجل عجوز يربط بيته ببالونات ويطير\", \"Up\", \"عربي/مشهور\"],\n [\"فلم عن رجل يعيش نفس اليوم مرة ورا مرة\", \"Groundhog Day\", \"عربي/مشهور\"],\n [\"مسلسل عن مدرس كيمياء يصير يصنع مخدرات\", \"Breaking Bad\", \"عربي/مشهور\"],\n [\"فلم عن محامي يشتغل من سيارته\", \"The Lincoln Lawyer\", \"عربي/مشهور\"],\n [\"فلم عن عالم افتراضي والبطل ياخذ حبة حمرا\", \"The Matrix\", \"عربي/مشهور\"],\n [\"فلم عن سفر بالفضاء عبر ثقب دودي عشان ينقذون البشرية\", \"Interstellar\", \"عربي/مشهور\"],\n [\"فلم كوري عن عائلة فقيرة تشتغل عند عائلة غنية\", \"Parasite\", \"عربي/مشهور\"],\n [\"مسلسل عن انفجار مفاعل نووي بالاتحاد السوفيتي\", \"Chernobyl\", \"عربي/مشهور\"],\n [\"فلم عن سفينة تضرب جبل جليد وتغرق\", \"Titanic\", \"عربي/مشهور\"],\n /* ═══ عربي · دفعة جديدة ═══ */\n [\"مسلسل عن ملكة بريطانيا من شبابها لين كبرت\", \"The Crown\", \"عربي/جديد\"],\n [\"فلم عن مهرج يصير مجرم بمدينة فاسدة\", \"Joker\", \"عربي/جديد\"],\n [\"مسلسل عن أطفال يلاقون بنت بقوى خارقة ووحش من عالم ثاني\", \"Stranger Things\", \"عربي/جديد\"],\n [\"فلم عن رجل محبوس بالسجن سنين وحفر نفق للهروب\", \"The Shawshank Redemption\", \"عربي/جديد\"],\n [\"فلم عن ملاكم يتدرب على درج فيلادلفيا\", \"Rocky\", \"عربي/جديد\"],\n [\"مسلسل عن عائلات تتقاتل على عرش والتنانين\", \"Game of Thrones\", \"عربي/جديد\"],\n [\"فلم عن رجل عنده ذاكرة قصيرة يكتب على جسمه\", \"Memento\", \"عربي/جديد\"],\n [\"فلم عن أب يدور ابنه بمباراة كرة بإيطاليا بعد الحرب\", \"Bicycle Thieves\", \"عربي/غامض\"],\n [\"فلم عن سمكة صغيرة ضاعت وأبوها يدورها بالبحر\", \"Finding Nemo\", \"عربي/جديد\"],\n [\"فلم عن رجل يركض بأمريكا كلها وله حياة غريبة\", \"Forrest Gump\", \"عربي/جديد\"],\n [\"مسلسل عن ست نفر أصدقاء بشقة بنيويورك\", \"Friends\", \"عربي/جديد\"],\n [\"فلم عن عصابة تسرق بنك وكلهم لابسين أقنعة\", \"Money Heist\", \"عربي/جديد\"],\n [\"فلم رعب عن فندق وكاتب يجن بالشتاء\", \"The Shining\", \"عربي/جديد\"],\n [\"فلم عن سيارة زمن وولد يرجع للماضي يقابل أهله\", \"Back to the Future\", \"عربي/جديد\"],\n [\"فلم عن كائن فضائي صغير يصادق ولد ويبي يرجع بيته\", \"E.T. the Extra-Terrestrial\", \"عربي/جديد\"],\n /* ═══ إنجليزي ═══ */\n [\"a man is stranded alone on mars and grows potatoes\", \"The Martian\", \"إنجليزي\"],\n [\"a shark terrorizes a beach town and three men hunt it\", \"Jaws\", \"إنجليزي\"],\n [\"toys come alive when humans are not looking\", \"Toy Story\", \"إنجليزي\"],\n [\"a rat wants to become a chef in paris\", \"Ratatouille\", \"إنجليزي\"],\n [\"a woman kills bill after waking from a coma\", \"Kill Bill: Vol. 1\", \"إنجليزي\"],\n [\"gladiator seeks revenge against the emperor who killed his family\", \"Gladiator\", \"إنجليزي\"],\n [\"a hobbit carries a ring to a volcano to destroy it\", \"The Lord of the Rings: The Fellowship of the Ring\", \"إنجليزي\"],\n [\"a submarine crew during world war two under pressure\", \"Das Boot\", \"إنجليزي/غامض\"],\n [\"an old man and a boy tiger stuck on a lifeboat\", \"Life of Pi\", \"إنجليزي\"],\n [\"a bus that explodes if it drives under fifty miles per hour\", \"Speed\", \"إنجليزي\"],\n /* ═══ وصف غامض/ناقص — الاختبار الأصعب ═══ */\n [\"فلم عن رجل يلبس بدلة خفافيش ويحارب الجريمة\", \"The Dark Knight\", \"غامض\"],\n [\"مسلسل عن أطباء بمستشفى وقصص حب بينهم\", \"Grey's Anatomy\", \"غامض\"],\n [\"فلم عن حرب النجوم والسيوف الضوئية\", \"Star Wars\", \"غامض\"],\n [\"فلم عن ديناصورات ترجع للحياة بجزيرة\", \"Jurassic Park\", \"غامض\"],\n [\"فلم عن رجل أخضر كبير يزعل ويكسر كل شي\", \"The Incredible Hulk\", \"غامض\"],\n];\n\nconst tok = (s: string) => (s || \"\").toLowerCase().match(/[a-z0-9]+/g) || [];\n\nfunction buildBM25(docs: string[][]) {\n const N = docs.length;\n const df = new Map();\n for (const d of docs) for (const w of new Set(d)) df.set(w, (df.get(w) || 0) + 1);\n const len = docs.map(d => d.length);\n const avg = len.reduce((a, b) => a + b, 0) / N;\n const tf = docs.map(d => { const m = new Map(); for (const w of d) m.set(w, (m.get(w) || 0) + 1); return m; });\n const k1 = 1.5, b = 0.75;\n return (q: string[]) => {\n const s = new Float64Array(N);\n for (const w of q) {\n const n = df.get(w); if (!n) continue;\n const idf = Math.log(1 + (N - n + 0.5) / (n + 0.5));\n for (let i = 0; i < N; i++) {\n const f = tf[i].get(w); if (!f) continue;\n s[i] += idf * (f * (k1 + 1)) / (f + k1 * (1 - b + b * len[i] / avg));\n }\n }\n return s;\n };\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"45\");\n\n const C: any[] = (await blob.getJSON(\"imdbcorpus_v1\")) as any[] || [];\n const titles = C.map(r => String(r.title || \"\"));\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const bmT = buildBM25(C.map(r => tok(r.ov)));\n const bmI = buildBM25(C.map(r => tok((r.plots || []).join(\" \"))));\n const bmB = buildBM25(C.map(r => tok(r.ov + \" \" + (r.plots || []).join(\" \"))));\n\n const slice = CASES.slice(from, to);\n const rows: any[] = [];\n const tally: any = {};\n const t0 = Date.now();\n\n for (const [ar, exp, tag] of slice) {\n let en = ar;\n if (/[\\u0600-\\u06FF]/.test(ar)) {\n try {\n const t = await fetch(`${SH}/translate?q=${encodeURIComponent(ar)}&target=en`, { signal: AbortSignal.timeout(15000) })\n .then(r => r.ok ? r.json() : null);\n if (t?.text) en = String(t.text);\n } catch { /* نكمل */ }\n }\n const q = tok(en);\n const row: any = { exp, tag };\n for (const [name, fn] of [[\"tmdb\", bmT], [\"imdb\", bmI], [\"both\", bmB]] as [string, any][]) {\n const s = fn(q);\n const order = Array.from(s.keys()).sort((a: number, b: number) => s[b] - s[a]);\n /* مطابقة مرنة: العنوان المتوقَّع جزء من عنوان النتيجة أو العكس */\n const want = norm(exp);\n const rank = order.findIndex((i: number) => {\n const t = norm(titles[i]);\n return t === want || t.startsWith(want) || want.startsWith(t);\n });\n row[name] = rank;\n if (name === \"both\") row.top = titles[order[0]];\n tally[name] = tally[name] || [0, 0, 0, 0];\n if (rank === 0) tally[name][0]++;\n if (rank >= 0 && rank < 5) tally[name][1]++;\n if (rank >= 0 && rank < 20) tally[name][2]++;\n tally[name][3]++;\n }\n rows.push(row);\n }\n\n const byTag: any = {};\n for (const r of rows) {\n byTag[r.tag] = byTag[r.tag] || { n: 0, first: 0, top5: 0 };\n byTag[r.tag].n++;\n if (r.imdb === 0) byTag[r.tag].first++;\n if (r.imdb >= 0 && r.imdb < 5) byTag[r.tag].top5++;\n }\n\n const pct = (a: number, b: number) => Math.round(a / b * 100) + \"٪\";\n return Response.json({\n corpus: C.length, tested: rows.length, ms: Date.now() - t0,\n summary: Object.fromEntries(Object.entries(tally).map(([k, v]: any) =>\n [k, `أول ${v[0]}/${v[3]} (${pct(v[0], v[3])}) · ضمن٥ ${v[1]}/${v[3]} (${pct(v[1], v[3])}) · ضمن٢٠ ${v[2]}/${v[3]}`])),\n byTag: Object.fromEntries(Object.entries(byTag).map(([k, v]: any) =>\n [k, `أول ${v.first}/${v.n} · ضمن٥ ${v.top5}/${v.n}`])),\n fails: rows.filter(r => r.imdb !== 0).map(r => `${r.exp}: IMDb ${r.imdb < 0 ? \"خارج\" : \"#\" + (r.imdb + 1)} ← ${r.top}`),\n });\n}\n",[],[841],"28e30795-0435-4a27-9059-da85b75f9215","best.ts","05d3c4d8-b081-11f1-aacd-1607ee4eb77e",110,["D",1789420362005],"/* أفضل ترتيب من الاستعلامين — لا متوسط ولا اختيار.\n كل عمل يأخذ أعلى مرتبة حققها بأيٍّ من الصيغتين، فلا يضيع مكسب أيٍّ منهما. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been being am \" +\n \"about after before while during into onto over under out up down off through across against between among \" +\n \"who whom whose that this these those which what when where how why there here \" +\n \"he she it they them him her his hers its their theirs we us our ours you your i me my mine \" +\n \"not no nor none any some all both each few more most other such only own same so than too very \" +\n \"can will just should now does do did done has have had having \" +\n \"movie film series show season episode\").split(\" \"));\n\nasync function q(terms: string[], lim = 30) {\n if (!terms.length) return [];\n const r = await sqlite.execute({\n sql: `SELECT title FROM plot5 WHERE plot5 MATCH ? ORDER BY bm25(plot5,1,1,1) LIMIT ${lim}`,\n args: [terms.map(w => `\"${w}\"`).join(\" OR \")],\n });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\n/* أفضل مرتبة: العمل يُقيَّم بأقوى ظهور له، والتعادل يُحسم بمجموع المرتبتين */\nfunction bestRank(A: string[], B: string[]) {\n const m = new Map();\n const add = (l: string[]) => l.forEach((t, i) => {\n const cur = m.get(t) || { best: 999, sum: 0 };\n cur.best = Math.min(cur.best, i + 1);\n cur.sum += i + 1;\n m.set(t, cur);\n });\n add(A); add(B);\n /* عمل ظهر بقائمة واحدة فقط: مجموعه يُعامَل كأنه ٩٩٩ بالقائمة الغائبة */\n for (const [t, v] of m) if (!(A.includes(t) && B.includes(t))) v.sum += 40;\n return [...m.entries()].sort((a, b) => (a[1].best - b[1].best) || (a[1].sum - b[1].sum)).map(([t]) => t);\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"15\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"مع الوظيفية\", \"بلا وظيفية\", \"أفضل الاثنين\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n let ms = 0;\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const all = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const clean = all.filter(w => !STOP.has(w));\n const use = clean.length >= 2 ? clean : all;\n\n const t0 = Date.now();\n const A = await q(all), B = await q(use);\n const C = bestRank(A, B);\n ms += Date.now() - t0;\n\n const lists: Record = { \"مع الوظيفية\": A, \"بلا وظيفية\": B, \"أفضل الاثنين\": C };\n const want = norm(exp);\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.join(\" · \"),\n زمن: Math.round(ms / n) + \"ms للبحث\",\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[849],"35e9ae69-12fa-405c-8206-acef9afc49f6","bench.ts","a5c06a0e-af9c-11f1-8730-1607ee4eb77e",12,["D",1789322312741],"import { findByPlot } from \"./engine.ts\";\n\nconst TESTS = [\n \"فلم عن طبيب يخبي زوجته الميتة بالقبو\",\n \"مسلسل كوري عن لعبة موت والفايز ياخذ فلوس\",\n \"فلم قديم عن رجل يعيش نفس اليوم مرة ورا مرة\",\n \"مسلسل عن محامي يشتغل بسيارته\",\n \"فلم عن ولد يقدر يشوف الأموات\",\n \"مسلسل عن شركة موظفينها يفصلون ذاكرة الشغل عن حياتهم\",\n \"فلم عن رجل عجوز يربط بيته ببالونات ويطير\",\n \"فلم عن غواصة روسية تهرب لأمريكا\",\n];\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n const i = Number(url.searchParams.get(\"i\") || \"0\");\n const slice = TESTS.slice(i, i + 3);\n const out: any[] = [];\n for (const t of slice) {\n try {\n const r = await findByPlot(t);\n out.push({\n q: t,\n tr: r.debug.translated,\n n: r.debug.resultCount,\n ms: r.debug.ms,\n top: (r.items || []).map((x: any) => x.title + (x.year ? \" (\" + x.year + \")\" : \"\")),\n });\n } catch (e) {\n out.push({ q: t, error: String(e).slice(0, 200) });\n }\n }\n return Response.json(out);\n}\n",[],[857],"e066c9b4-e363-4292-bd52-6896283c689f","autopsy.ts","9f4c10d0-b080-11f1-83fb-1607ee4eb77e",118,["D",1789442894572],"/* تشريح الفشل الثلاثة على القاعدة الكاملة — هل الكلمة المتذكَّرة موجودة بالنص أصلاً؟ */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst T: [string, string, string[]][] = [\n [\"Forrest Gump\", \"فلم عن رجل يركض بأمريكا كلها وله حياة غريبة\", [\"run\", \"running\", \"america\", \"jog\", \"cross country\"]],\n [\"Back to the Future\", \"فلم عن سيارة سيارة زمن وولد يرجع للماضي يقابل أهله\", [\"time machine\", \"delorean\", \"1955\", \"past\", \"parents\", \"car\"]],\n [\"Grey's Anatomy\", \"مسلسل عن أطباء بمستشفى وقصص حب بينهم\", [\"doctor\", \"surgeon\", \"hospital\", \"romance\", \"intern\"]],\n];\n\nexport default async function (req: Request): Promise {\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const out: any = {};\n\n for (const [title, ar, words] of T) {\n const o: any = { الترجمة: cache[ar] || ar };\n const r = await sqlite.execute({\n sql: `SELECT length(plots), length(kw), substr(plots,1,260) FROM plot5 WHERE title = ? LIMIT 1`,\n args: [title],\n });\n if (!(r.rows || []).length) { o.الحالة = \"❌ غير موجود\"; out[title] = o; continue; }\n const row: any = r.rows[0];\n o.طول_الملخصات = row[0]; o.طول_الوسوم = row[1]; o.بداية_النص = row[2];\n\n o.الكلمات_المتذكَّرة = {};\n for (const w of words) {\n const q = await sqlite.execute({\n sql: `SELECT (CASE WHEN lower(plots) LIKE ? THEN 'ملخّص' ELSE '' END) ||\n (CASE WHEN lower(kw) LIKE ? THEN ' وسوم' ELSE '' END)\n FROM plot5 WHERE title = ? LIMIT 1`,\n args: [\"%\" + w + \"%\", \"%\" + w + \"%\", title],\n });\n const v = String(q.rows[0][0] || \"\").trim();\n o.الكلمات_المتذكَّرة[w] = v || \"❌ غير موجودة إطلاقاً\";\n }\n out[title] = o;\n }\n return Response.json(out);\n}\n",[],[865],"e92f1969-1353-4a88-8613-731661e1ed15","assoc.ts","01537b3c-b073-11f1-93a7-1607ee4eb77e",91,["D",1789414341502],"/* بناء جدول توسيع من ملخّصاتنا نفسها:\n لكل كلمة، الكلمات التي تشاركها السياق (تظهر بنفس المستندات أكثر من الصدفة).\n المقياس: PMI معياري — يرفع \"thieves↔burglars\" ويخفض \"thieves↔the\".\n صفر مصدر خارجي · يُبنى مرة واحدة · يُخزَّن بالبلوب. */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst KEY = \"corpus_v2\";\nconst OUT = \"assoc_v1\";\n\n/* كلمات لا تصلح للتوسيع أبداً */\nconst STOP = new Set((\"a an the of in on at to for from by with and or but is are was were be been being \" +\n \"about after who that this these those its their his her he she it they them we you i not no \" +\n \"has have had do does did will would can could should may might must there here when where how why \" +\n \"what which while during into onto over under out up down off then than also very just only more most \" +\n \"one two three first second new old man woman boy girl people time day night year years life \" +\n \"film movie series show season episode story character characters plot scene\").split(\" \"));\n\nconst tok = (s: string) => (s || \"\").toLowerCase().match(/[a-z]{3,}/g) || [];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const minDf = Number(u.searchParams.get(\"mindf\") || \"8\");\n const maxDfRatio = Number(u.searchParams.get(\"maxdf\") || \"0.12\");\n const topK = Number(u.searchParams.get(\"k\") || \"6\");\n\n const C: any[] = (await blob.getJSON(KEY)) as any[] || [];\n const t0 = Date.now();\n\n /* ١) مجموعات كلمات لكل مستند (وجود فقط، لا تكرار) */\n const docs: Set[] = C.map(r => {\n const s = new Set();\n for (const w of tok(String(r.ov || \"\") + \" \" + (r.plots || []).join(\" \"))) if (!STOP.has(w)) s.add(w);\n return s;\n });\n const N = docs.length;\n\n /* ٢) تكرار كل كلمة */\n const df = new Map();\n for (const d of docs) for (const w of d) df.set(w, (df.get(w) || 0) + 1);\n\n /* ٣) المفردات الصالحة: ليست نادرة جداً ولا عامة جداً */\n const vocab = [...df.entries()]\n .filter(([, n]) => n >= minDf && n <= N * maxDfRatio)\n .map(([w]) => w);\n const vset = new Set(vocab);\n const idx = new Map(vocab.map((w, i) => [w, i]));\n\n /* ٤) مصفوفة التشارك — نمر على المستندات مرة واحدة */\n const co = new Map>();\n for (const d of docs) {\n const ws = [...d].filter(w => vset.has(w)).map(w => idx.get(w)!);\n if (ws.length > 120) continue; /* مستند طويل جداً يضخّم كل شيء */\n for (let i = 0; i < ws.length; i++) {\n let m = co.get(ws[i]); if (!m) { m = new Map(); co.set(ws[i], m); }\n for (let j = 0; j < ws.length; j++) if (i !== j) m.set(ws[j], (m.get(ws[j]) || 0) + 1);\n }\n }\n\n /* ٥) PMI معياري لكل زوج، ونحتفظ بأقوى K لكل كلمة */\n const dfArr = vocab.map(w => df.get(w)!);\n const table: Record = {};\n for (const [i, m] of co) {\n const pi = dfArr[i] / N;\n const scored: [number, number][] = [];\n for (const [j, c] of m) {\n if (c < 4) continue;\n const pj = dfArr[j] / N, pij = c / N;\n const pmi = Math.log(pij / (pi * pj));\n const npmi = pmi / -Math.log(pij); /* معياري بين -١ و١ */\n if (npmi > 0.22) scored.push([j, npmi]);\n }\n if (!scored.length) continue;\n scored.sort((a, b) => b[1] - a[1]);\n table[vocab[i]] = scored.slice(0, topK).map(([j]) => vocab[j]);\n }\n\n await blob.setJSON(OUT, table);\n const keys = Object.keys(table);\n return Response.json({\n ms: Date.now() - t0,\n مستندات: N, مفردات_صالحة: vocab.length, كلمات_لها_توسيع: keys.length,\n عينات: {\n thieves: table[\"thieves\"], burglars: table[\"burglars\"], burglar: table[\"burglar\"],\n balloons: table[\"balloons\"], silence: table[\"silence\"], pill: table[\"pill\"],\n cave: table[\"cave\"], dreams: table[\"dreams\"],\n },\n });\n}\n",[],[873],"9be85982-f753-4a44-95dd-88b82f396f23","arabic.ts","42fd42fe-b134-11f1-85cb-1607ee4eb77e",191,["D",1789497344543],"/* إصلاح ثغرة الأعمال العربية: سحب بلغة العمل الأصلية (ar) — غير مغطاة بالترتيب العالمي */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst SH = \"https://q8kamikaze--0265d7dc765711f1baf01607ee4eb77e.web.val.run\";\nconst GQL = \"https://api.graphql.imdb.com/\";\nconst H = { \"content-type\": \"application/json\", \"origin\": \"https://www.imdb.com\",\n \"referer\": \"https://www.imdb.com/\",\n \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120 Safari/537.36\" };\n\nasync function j(u: string) {\n try { const r = await fetch(u, { signal: AbortSignal.timeout(15000) }); return r.ok ? await r.json() : null; }\n catch { return null; }\n}\nasync function gql(q: string) {\n try {\n const r = await fetch(GQL, { method: \"POST\", headers: H, body: JSON.stringify({ query: q }), signal: AbortSignal.timeout(20000) });\n return await r.json();\n } catch { return null; }\n}\nasync function pool(items: T[], n: number, fn: (x: T) => Promise) {\n const out: R[] = []; let i = 0;\n await Promise.all(Array.from({ length: n }, async () => {\n while (i < items.length) { const k = i++; out[k] = await fn(items[k]); }\n }));\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"1\");\n const to = Number(u.searchParams.get(\"to\") || \"15\");\n const lang = u.searchParams.get(\"lang\") || \"ar\";\n\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n const seen = new Set(all.map((r: any) => r.kind + \":\" + r.id));\n\n const pages: [string, number][] = [];\n for (let p = from; p <= to; p++) { pages.push([\"movie\", p]); pages.push([\"tv\", p]); }\n\n /* الاسم العربي والإنجليزي معاً — المستخدم قد يكتب أيّهما */\n const base = (await pool(pages, 10, async ([kind, p]) => {\n const d = await j(`${SH}/tmdb/discover/${kind}?with_original_language=${lang}&sort_by=vote_count.desc&page=${p}&vote_count.gte=5`);\n return (d?.results || []).map((x: any) => ({\n id: x.id, kind,\n title: x.title || x.name,\n titleAr: x.original_title || x.original_name || \"\",\n year: String(x.release_date || x.first_air_date || \"\").slice(0, 4),\n ov: String(x.overview || \"\").slice(0, 400),\n votes: x.vote_count || 0,\n }));\n })).flat().filter((r: any) => !seen.has(r.kind + \":\" + r.id));\n\n const withImdb = await pool(base, 14, async (r: any) => {\n const e = await j(`${SH}/tmdb/${r.kind}/${r.id}/external_ids`);\n r.imdb = e?.imdb_id || null;\n return r;\n });\n const ok = withImdb.filter((r: any) => r.imdb);\n\n /* ملخّصات IMDb لمن له معرّف */\n const groups: any[][] = [];\n for (let i = 0; i < ok.length; i += 5) groups.push(ok.slice(i, i + 5));\n await pool(groups, 6, async (g: any[]) => {\n const alias = g.map((r, i) => `t${i}: title(id: ${JSON.stringify(r.imdb)}) {\n plots(first: 6) { edges { node { plotText { plainText } } } } }`).join(\" \");\n const res = await gql(`query { ${alias} }`);\n g.forEach((r, i) => {\n r.plots = ((res?.data?.[`t${i}`]?.plots?.edges) || [])\n .map((e: any) => String(e.node?.plotText?.plainText || \"\")).filter(Boolean);\n });\n return g;\n });\n\n /* من بلا معرّف IMDb نبقيه بملخّص TMDB — أفضل من إسقاطه */\n const noImdb = withImdb.filter((r: any) => !r.imdb && r.ov);\n const merged = all.concat(ok, noImdb);\n await blob.setJSON(\"corpus_v2\", merged);\n\n const arabic = merged.filter((r: any) => r.titleAr && /[\\u0600-\\u06FF]/.test(r.titleAr)).length;\n return Response.json({\n \"أضيف\": ok.length + noImdb.length, \"منها بمعرّف IMDb\": ok.length,\n \"إجمالي القاعدة\": merged.length, \"أعمال بعنوان عربي\": arabic,\n عينة: ok.slice(0, 5).map((r: any) => `${r.title} / ${r.titleAr} (${r.year})`),\n });\n}\n",[],[881],"338181d3-5042-45fa-8dac-3fd7b010d4a3","append.ts","04329b4c-b0c5-11f1-aef1-1607ee4eb77e",151,["D",1789449565055],"/* تصميم «الإضافة لا المزاحمة»:\n الفهرس الأساسي (ملخّص+وسوم) يحكم الترتيب كما هو — بلا أي مساس.\n المصادر الغنية (حبكة ويكيبيديا + حلقات) تُستعمل فقط لإضافة أعمال **غائبة**\n عن النتائج، تُلحَق بالذيل. فلا يمكنها زحزحة نتيجة صحيحة أبداً. */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nasync function q(sql: string, match: string) {\n const r = await sqlite.execute({ sql, args: [match] });\n return (r.rows || []).map((x: any) => String(x[0]));\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"15\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const V = [\"أساسي وحده\", \"إضافة لا مزاحمة\", \"دمج تقليدي (وزن ١)\"];\n const tally: any = {}; for (const v of V) tally[v] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n\n /* ١) الأساسي: ملخّصات + وسوم — هو من يحكم الترتيب */\n const base = await q(\n `SELECT title FROM plot5 WHERE plot5 MATCH ? ORDER BY bm25(plot5,1,1,1) LIMIT 20`, match);\n\n /* ٢) الغني: حبكة ويكيبيديا + الحلقات وحدها */\n const rich = await q(\n `SELECT title FROM plot8 WHERE plot8 MATCH ? ORDER BY bm25(plot8,1,0,0,1) LIMIT 20`, match);\n\n /* الإلحاق: ما لم يظهر بالأساسي يُضاف بالذيل — صفر إزاحة */\n const seen = new Set(base);\n const appended = [...base, ...rich.filter(t => !seen.has(t))].slice(0, 20);\n\n /* ٣) للمقارنة: الدمج التقليدي داخل الحساب نفسه */\n const merged = await q(\n `SELECT title FROM plot8 WHERE plot8 MATCH ? ORDER BY bm25(plot8,1,1,1,1) LIMIT 20`, match);\n\n const lists: Record = {\n \"أساسي وحده\": base, \"إضافة لا مزاحمة\": appended, \"دمج تقليدي (وزن ١)\": merged,\n };\n const want = norm(exp);\n const parts: string[] = [];\n for (const v of V) {\n const rk = lists[v].findIndex(t => { const n = norm(t); return n === want || n.startsWith(want) || want.startsWith(n); });\n if (rk === 0) tally[v][0]++;\n if (rk >= 0 && rk < 5) tally[v][1]++;\n if (rk >= 0 && rk < 20) tally[v][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.join(\" · \"),\n النتيجة: Object.fromEntries(V.map(v => [v, `أول ${tally[v][0]}/${n} · ضمن٥ ${tally[v][1]}/${n} · ضمن٢٠ ${tally[v][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[889],"7d138a1b-e50b-45c2-a89c-5bd2ff0c5da6","adpull.ts","19347ca4-b138-11f1-9bd6-1607ee4eb77e",207,["D",1789498992417],"/* تنزيل ودمج بيانات الوصف السمعي (AutoAD/CMD) — ١٠١ ألف جملة وصف مشهد */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst URL_CSV = \"http://www.robots.ox.ac.uk/~htd/ad3/cmd_ad_anno_v1.csv\";\nconst UA = { \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/120 Safari/537.36\" };\n\n/* محلّل CSV بسيط يحترم علامات الاقتباس — الوصف قد يحوي فواصل */\nfunction parseLine(line: string): string[] {\n const out: string[] = []; let cur = \"\", q = false;\n for (let i = 0; i < line.length; i++) {\n const c = line[i];\n if (c === '\"') { if (q && line[i + 1] === '\"') { cur += '\"'; i++; } else q = !q; }\n else if (c === \",\" && !q) { out.push(cur); cur = \"\"; }\n else cur += c;\n }\n out.push(cur);\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const out: any = {};\n\n if (u.searchParams.get(\"stage\") === \"check\") {\n const idx: any = await blob.getJSON(\"ad_index\");\n const keys = idx ? Object.keys(idx) : [];\n return Response.json({\n \"أفلام لها وصف سمعي\": keys.length,\n \"متوسط طول الوصف\": keys.length ? Math.round(keys.reduce((a, k) => a + idx[k].length, 0) / keys.length) + \" حرف\" : \"—\",\n عينة: keys.slice(0, 3).map(k => k + \": \" + String(idx[k]).slice(0, 180)),\n });\n }\n\n const t0 = Date.now();\n const r = await fetch(URL_CSV, { headers: UA, signal: AbortSignal.timeout(50000) });\n const csv = await r.text();\n out.status = r.status; out[\"حجم\"] = csv.length;\n\n const lines = csv.split(\"\\n\");\n const head = parseLine(lines[0]);\n const iText = head.indexOf(\"text\"), iImdb = head.indexOf(\"imdbid\"), iTitle = head.indexOf(\"movie_title\");\n out.أعمدة = { iText, iImdb, iTitle };\n\n /* نجمع جمل كل فلم بمعرّف IMDb */\n const byId: Record = {};\n const titles: Record = {};\n for (let i = 1; i < lines.length; i++) {\n const c = parseLine(lines[i]);\n const id = c[iImdb]?.trim();\n const txt = c[iText]?.trim();\n if (!id || !txt || txt.length < 8) continue;\n (byId[id] ||= []).push(txt);\n if (!titles[id]) titles[id] = c[iTitle] || \"\";\n }\n\n /* نصّ واحد لكل فلم — نحدّه لتفادي التمييع الذي قِسناه مراراً */\n const index: Record = {};\n for (const [id, arr] of Object.entries(byId)) index[id] = arr.join(\" \").slice(0, 8000);\n\n await blob.setJSON(\"ad_index\", index);\n await blob.setJSON(\"ad_titles\", titles);\n\n const ks = Object.keys(index);\n out.ms = Date.now() - t0;\n out[\"أفلام مغطّاة\"] = ks.length;\n out[\"إجمالي الجمل\"] = Object.values(byId).reduce((a, b) => a + b.length, 0);\n out[\"متوسط الجمل لكل فلم\"] = Math.round(Object.values(byId).reduce((a, b) => a + b.length, 0) / ks.length);\n out[\"متوسط طول النص\"] = Math.round(ks.reduce((a, k) => a + index[k].length, 0) / ks.length) + \" حرف\";\n out.عينة = ks.slice(0, 3).map(k => `${titles[k]} (${k}): ${index[k].slice(0, 140)}`);\n return Response.json(out);\n}\n",[],[897],"88ac9819-f995-42f9-9050-3f12b8619a64","admerge2.ts","b0e97ae4-b13e-11f1-9784-1607ee4eb77e",216,["D",1789501823918],"/* ربط معرّفات AudioVault بـIMDb ثم الدمج بالقاعدة */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nconst UA = { \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/120 Safari/537.36\" };\nconst MAP_URL = \"http://www.robots.ox.ac.uk/~htd/autoad/audiovault_imdbids.csv\";\n\nexport default async function (req: Request): Promise {\n const out: any = {};\n\n /* ١) جدول الربط: avid → imdbid */\n const r = await fetch(MAP_URL, { headers: UA, signal: AbortSignal.timeout(30000) });\n const csv = await r.text();\n const lines = csv.split(\"\\n\");\n const map: Record = {};\n let exact = 0;\n for (let i = 1; i < lines.length; i++) {\n const c = lines[i].split(\",\");\n const avid = (c[0] || \"\").trim(), imdb = (c[1] || \"\").trim(), ok = (c[2] || \"\").trim();\n if (!avid || !imdb) continue;\n /* نأخذ المطابقات المؤكَّدة فقط — التقريبية تُدخل أفلاماً خاطئة */\n if (ok !== \"True\") continue;\n map[avid] = imdb; exact++;\n }\n out[\"جدول الربط\"] = { \"أسطر\": lines.length, \"مطابقات مؤكَّدة\": exact };\n\n /* ٢) الوصف السمعي المجمَّع */\n const st: any = (await blob.getJSON(\"ad8k_state\")) || { movies: {} };\n const movies: Record = st.movies || {};\n out[\"أعمال بوصف سمعي\"] = Object.keys(movies).length;\n\n /* ٣) فهرس نهائي: imdbid → نص الوصف */\n const byImdb: Record = {};\n for (const [avid, txt] of Object.entries(movies)) {\n const imdb = map[avid];\n if (imdb) byImdb[imdb] = txt;\n }\n out[\"مربوط بمعرّف IMDb\"] = Object.keys(byImdb).length;\n\n /* ٤) الدمج بالقاعدة */\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n let hit = 0, chars = 0;\n const sample: string[] = [];\n for (const rec of all) {\n const id = String(rec.imdb || \"\");\n if (id && byImdb[id]) {\n rec.ad = byImdb[id];\n hit++; chars += byImdb[id].length;\n if (sample.length < 8) sample.push(rec.title + \" (\" + rec.year + \")\");\n }\n }\n await blob.setJSON(\"corpus_v2\", all);\n\n out[\"مطابق بقاعدتنا\"] = hit + \" (\" + Math.round(hit / all.length * 100) + \"٪ من \" + all.length + \")\";\n out[\"متوسط الطول\"] = hit ? Math.round(chars / hit) + \" حرف\" : \"—\";\n out.عينة = sample;\n return Response.json(out);\n}\n",[],[905],"60bc7111-96bd-4329-a84d-337c5fdde5a4","admerge.ts","23d6e728-b138-11f1-8d3d-1607ee4eb77e",208,["D",1789499010265],"/* دمج الوصف السمعي بالقاعدة — الربط بمعرّف IMDb (دقيق ١٠٠٪ بلا تخمين أسماء) */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nexport default async function (req: Request): Promise {\n const idx: Record = (await blob.getJSON(\"ad_index\")) as any || {};\n const all: any[] = (await blob.getJSON(\"corpus_v2\")) as any[] || [];\n\n let hit = 0, chars = 0;\n const matched: string[] = [];\n for (const r of all) {\n const id = String(r.imdb || \"\");\n if (id && idx[id]) {\n r.ad = idx[id];\n hit++; chars += idx[id].length;\n if (matched.length < 8) matched.push(r.title + \" (\" + r.year + \")\");\n }\n }\n await blob.setJSON(\"corpus_v2\", all);\n\n return Response.json({\n \"أفلام الوصف السمعي\": Object.keys(idx).length,\n \"مطابق بقاعدتنا\": hit + \" (\" + Math.round(hit / Object.keys(idx).length * 100) + \"٪ من المجموعة)\",\n \"من قاعدتنا\": Math.round(hit / all.length * 100) + \"٪ من \" + all.length,\n \"متوسط الطول\": hit ? Math.round(chars / hit) + \" حرف\" : \"—\",\n عينة: matched,\n });\n}\n",[],[913],"ed10e24d-e03f-4461-b385-eb04140134a6","adbench.ts","477e94d2-b142-11f1-96a3-1607ee4eb77e",218,["D",1789503365047],"/* القياس الحاسم: الفهرس السداسي (بالوصف السمعي) مقابل الخماسي — ٤٥ حالة، النطاق الكامل */\nimport { sqlite } from \"https://esm.town/v/stevekrouse/sqlite\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\nimport { CASES } from \"./cases.ts\";\n\nconst V: [string, string, string][] = [\n [\"خماسي (بلا وصف)\", \"plotP\", \"bm25(plotP,1,1,1,2,2)\"],\n [\"سداسي وصف ٠٫٣\", \"plotC\", \"bm25(plotC,1,1,1,2,2,0.3)\"],\n [\"سداسي وصف ٠٫٧\", \"plotC\", \"bm25(plotC,1,1,1,2,2,0.7)\"],\n [\"سداسي وصف ١٫٢\", \"plotC\", \"bm25(plotC,1,1,1,2,2,1.2)\"],\n];\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const from = Number(u.searchParams.get(\"from\") || \"0\");\n const to = Number(u.searchParams.get(\"to\") || \"23\");\n const cache: any = (await blob.getJSON(\"trcache_v1\")) || {};\n const norm = (s: string) => s.toLowerCase().replace(/[^a-z0-9]+/g, \" \").trim();\n\n const tally: any = {}; for (const [n] of V) tally[n] = [0, 0, 0];\n const rows: string[] = [];\n\n for (const [ar, exp] of CASES.slice(from, to)) {\n const en = cache[ar] || ar;\n const terms = [...new Set((en.toLowerCase().match(/[a-z0-9]+/g) || []).filter((w: string) => w.length > 1))] as string[];\n const match = terms.map(w => `\"${w}\"`).join(\" OR \");\n const want = norm(exp);\n const parts: string[] = [];\n\n for (const [name, table, rank] of V) {\n const r = await sqlite.execute({\n sql: `SELECT title FROM ${table} WHERE ${table} MATCH ? ORDER BY ${rank} LIMIT 20`, args: [match] });\n const l = (r.rows || []).map((x: any) => String(x[0]));\n const rk = l.findIndex(t => { const x = norm(t); return x === want || x.startsWith(want) || want.startsWith(x); });\n if (rk === 0) tally[name][0]++;\n if (rk >= 0 && rk < 5) tally[name][1]++;\n if (rk >= 0 && rk < 20) tally[name][2]++;\n parts.push(rk < 0 ? \"خارج\" : String(rk + 1));\n }\n rows.push(`${exp}: ${parts.join(\" \")}`);\n }\n\n const n = rows.length;\n return Response.json({\n أعمدة: V.map(v => v[0]).join(\" · \"),\n النتيجة: Object.fromEntries(V.map(([name]) =>\n [name, `أول ${tally[name][0]}/${n} · ضمن٥ ${tally[name][1]}/${n} · ضمن٢٠ ${tally[name][2]}/${n}`])),\n التفصيل: rows,\n });\n}\n",[],[921],"b4d6b3ce-d481-4a32-9d7b-bd1f8291472b","ad8k.ts","b0536178-b13c-11f1-9297-1607ee4eb77e",215,["D",1789500963952],"/* تنزيل ملف الوصف السمعي الكبير (٣٢٧ ميجا) مجزّأً بنطاقات بايتات\n ثم بناء فهرس: معرّف AudioVault → نصّ الوصف، ولاحقاً → معرّف IMDb */\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\nconst URL_AD = \"http://www.robots.ox.ac.uk/~htd/autoad/ad-movie8k-v3-nomadnolsmdc_NSSD_jaesung_0p95_minAD100_pronthresh0p05.csv\";\nconst UA = { \"user-agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 Chrome/120 Safari/537.36\" };\nconst CHUNK = 24 * 1024 * 1024; /* ٢٤ ميجا بكل جولة — يناسب مهلة الطلب */\n\nfunction parseLine(line: string): string[] {\n const out: string[] = []; let cur = \"\", q = false;\n for (let i = 0; i < line.length; i++) {\n const c = line[i];\n if (c === '\"') { if (q && line[i + 1] === '\"') { cur += '\"'; i++; } else q = !q; }\n else if (c === \",\" && !q) { out.push(cur); cur = \"\"; }\n else cur += c;\n }\n out.push(cur);\n return out;\n}\n\nexport default async function (req: Request): Promise {\n const u = new URL(req.url);\n const part = Number(u.searchParams.get(\"part\") || \"0\");\n const out: any = { part };\n\n if (u.searchParams.get(\"stage\") === \"check\") {\n const st: any = (await blob.getJSON(\"ad8k_state\")) || { parts: {}, movies: {} };\n return Response.json({\n \"أجزاء منزَّلة\": Object.keys(st.parts).length,\n \"أعمال مجمَّعة\": Object.keys(st.movies || {}).length,\n \"آخر جزء\": st.last,\n });\n }\n\n const st: any = (await blob.getJSON(\"ad8k_state\")) || { parts: {}, movies: {}, tail: \"\" };\n const start = part * CHUNK;\n const end = start + CHUNK - 1;\n\n const t0 = Date.now();\n const r = await fetch(URL_AD, { headers: { ...UA, range: `bytes=${start}-${end}` }, signal: AbortSignal.timeout(50000) });\n out.status = r.status;\n if (r.status !== 206 && r.status !== 200) return Response.json({ ...out, done: true });\n\n let text = await r.text();\n out[\"حجم الجزء\"] = text.length;\n\n /* السطر المقطوع من الجزء السابق يُلصق ببداية هذا الجزء */\n if (part > 0 && st.tail) text = st.tail + text;\n const lines = text.split(\"\\n\");\n st.tail = lines.pop() || \"\"; /* آخر سطر قد يكون ناقصاً */\n\n /* الأعمدة: text,start,end,speaker,speech_type,movie */\n let added = 0;\n for (const line of lines) {\n if (!line || line.startsWith(\"text,\")) continue;\n const c = parseLine(line);\n const txt = (c[0] || \"\").trim().replace(/^\"|\"$/g, \"\");\n const type = (c[4] || \"\").trim();\n const mv = (c[5] || \"\").trim();\n if (type !== \"ad\" || !mv || txt.length < 8) continue; /* الوصف فقط لا الحوار */\n const cur = st.movies[mv] || \"\";\n if (cur.length < 8000) { st.movies[mv] = cur + (cur ? \" \" : \"\") + txt; added++; }\n }\n\n st.parts[part] = added;\n st.last = part;\n await blob.setJSON(\"ad8k_state\", st);\n\n out.ms = Date.now() - t0;\n out[\"جمل أُضيفت\"] = added;\n out[\"أعمال حتى الآن\"] = Object.keys(st.movies).length;\n out[\"الجزء التالي\"] = part + 1;\n return Response.json(out);\n}\n",[],[929],["D",1789320435960],{"_11":12,"_130":71},{"_11":12,"_125":126,"_70":71,"_30":69,"_127":128,"_129":24},{"_39":116,"_13":16,"_33":36,"_117":80,"_118":-5,"_119":-5,"_120":-5,"_121":-5,"_122":-5,"_123":14,"_124":12},"ctx",{"_54":-5,"_149":973,"_974":975,"_976":977,"_978":24,"_979":24,"_980":24,"_981":93,"_982":-5,"_983":-7,"_984":-7,"_985":26,"_986":24,"_987":24,"_988":-5,"_989":24,"_990":24,"_991":24,"_992":-5},"ENV",{"_944":945,"_946":947,"_948":16,"_949":950,"_951":952,"_953":954,"_955":956,"_957":958,"_959":960,"_961":962,"_963":964,"_965":966,"_967":968,"_969":970,"_971":972},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

History · index3.ts