[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_194":195,"_196":197},"routes/_app.x.$handle.$name",{"_16":17,"_20":21,"_83":65,"_84":85,"_63":-7,"_86":-5,"_13":87,"_88":89,"_11":90,"_91":92,"_93":94,"_95":-5,"_96":-5,"_97":-5,"_98":26,"_99":100,"_101":102,"_103":-5,"_104":-5,"_105":106},"routes/_app.x.$handle.$name.search.$(branch)",{"_11":12,"_13":14,"_15":-5,"_16":17},"files",[],"project",{"_18":19,"_20":21,"_22":23,"_24":-5,"_25":26,"_27":28,"_29":28,"_30":31,"_32":-5,"_33":34,"_35":-5,"_36":26,"_37":-5,"_38":-5,"_39":40,"_41":42,"_43":-5,"_44":-5,"_45":-5,"_46":-5,"_47":48,"_49":48,"_50":51,"_52":51,"_53":51,"_54":48,"_55":48,"_56":57,"_58":59,"_60":61,"_62":26},"query","handle","khawjaahmad","id","6bba5b2c-4d5d-11f1-ab91-ee650bb23af1","name","playwright-gen","createdAt",["D",1778519905986],"deletedAt","legacyEnvVars",false,"privacy","public","httpPrivacy","description","TinyFish → AI → Playwright test generator","imageUrl","userId","f404441a-b070-431b-b152-17a2208cfa70","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","018b67b5-3681-41ae-8e5a-cdcf7cd1218d","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_16":17,"_82":78},"org",{"_16":17,"_75":76,"_77":78,"_33":34,"_79":80,"_81":26},"branch",{"_63":64,"_20":65,"_18":66,"_67":19,"_68":-5,"_69":-5,"_70":-5,"_71":-5,"_72":-5,"_73":21,"_74":17},"isOwned","version",8,"main","6bbbce58-4d5d-11f1-ab91-ee650bb23af1","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","ahmad","avatar","https://imagedelivery.net/iHX6Ovru0O7AjmyT5yZRoA/4ec015ce-90e5-4eef-7eaf-21e3da8cec00/square512?exp=1789592399&sig=850afdd394985736ea1d53cc04be2e1647f2bb45fb81c86e41a0f4bcc6af8300","tier","free","canRestrictVals","profileImageUrl","branchName","path","","adoptOrgId",{"_18":19,"_20":21,"_22":190,"_24":-5,"_25":26,"_27":28,"_29":28,"_30":31,"_32":-5,"_33":34,"_35":-5,"_36":26,"_37":-5,"_38":-5,"_39":40,"_41":42,"_43":-5,"_44":-5,"_45":-5,"_46":-5,"_47":48,"_49":48,"_50":51,"_52":51,"_53":51,"_54":48,"_55":48,"_56":191,"_58":192,"_60":193,"_62":26},"tree",[169,170,171,172,173],[169,170,171,172,173],"counts",{"_159":40,"_160":134,"_161":40,"_162":163,"_164":165},"contents",[107,108,109,110,111],"file","upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles",true,"grants","orgs","customDomains",{},{"_18":152,"_20":153,"_114":154,"_116":-5,"_117":66,"_118":155,"_120":102,"_22":156,"_63":134,"_122":95,"_124":157,"_126":26,"_127":-5,"_84":158},{"_18":146,"_20":147,"_114":148,"_116":-5,"_117":66,"_118":64,"_120":102,"_22":149,"_63":134,"_122":123,"_124":150,"_126":26,"_127":-5,"_84":151},{"_18":139,"_20":140,"_114":141,"_116":-5,"_117":66,"_118":142,"_120":102,"_22":143,"_63":40,"_122":123,"_124":144,"_126":26,"_127":-5,"_84":145},{"_18":129,"_20":130,"_114":131,"_116":-5,"_117":66,"_118":132,"_120":102,"_22":133,"_63":134,"_122":135,"_124":136,"_126":26,"_127":137,"_84":138},{"_18":112,"_20":113,"_114":115,"_116":-5,"_117":66,"_118":119,"_120":102,"_22":121,"_63":40,"_122":123,"_124":125,"_126":26,"_127":-5,"_84":128},"b6184f6c-4d5d-11f1-821b-ee650bb23af1","tinyfish.ts","fileId","b618500c-4d5d-11f1-821b-ee650bb23af1","parentId","branchId","branchVersion",3,"latest",["D",1778520030745],"fileType","script","content","// Thin wrapper over the TinyFish Fetch API.\n// Docs: https://docs.tinyfish.ai/api-reference/fetch-and-extract-content-from-urls\n\nconst TINYFISH_FETCH_URL = \"https://api.fetch.tinyfish.ai/\";\n\nexport class TinyFishError extends Error {}\nexport class CloudflareChallengeError extends Error {}\n\ninterface FetchHtmlArgs {\n url: string;\n apiKey: string;\n timeoutMs: number;\n}\n\ninterface TinyFishResult {\n url: string;\n final_url?: string;\n title?: string;\n text?: string; // when format=html, this holds the HTML string\n format?: string;\n}\n\ninterface TinyFishResponse {\n results?: TinyFishResult[];\n errors?: { url: string; error: string }[];\n}\n\nexport async function fetchHtml({ url, apiKey, timeoutMs }: FetchHtmlArgs): Promise {\n const ctrl = new AbortController();\n const t = setTimeout(() => ctrl.abort(), timeoutMs);\n let res: Response;\n try {\n res = await fetch(TINYFISH_FETCH_URL, {\n method: \"POST\",\n headers: {\n \"content-type\": \"application/json\",\n \"x-api-key\": apiKey,\n },\n body: JSON.stringify({\n urls: [url],\n format: \"html\",\n include_html_head: false,\n links: false,\n image_links: false,\n }),\n signal: ctrl.signal,\n });\n } catch (e) {\n clearTimeout(t);\n if ((e as Error).name === \"AbortError\") {\n throw new TinyFishError(`TinyFish fetch timed out after ${timeoutMs}ms`);\n }\n throw new TinyFishError(`TinyFish fetch failed: ${String(e)}`);\n }\n clearTimeout(t);\n\n if (!res.ok) {\n const detail = await safeText(res);\n throw new TinyFishError(`TinyFish HTTP ${res.status}: ${detail.slice(0, 500)}`);\n }\n\n let payload: TinyFishResponse;\n try {\n payload = await res.json();\n } catch {\n throw new TinyFishError(\"TinyFish returned non-JSON\");\n }\n\n if (payload.errors && payload.errors.length && (!payload.results || !payload.results.length)) {\n throw new TinyFishError(`TinyFish error: ${payload.errors[0].error}`);\n }\n const first = payload.results?.[0];\n if (!first || typeof first.text !== \"string\" || !first.text) {\n throw new TinyFishError(\"TinyFish returned no HTML content\");\n }\n\n const html = first.text;\n if (looksLikeChallenge(html, first.title)) {\n throw new CloudflareChallengeError(\n \"Target served an anti-bot challenge page. Cleaned HTML would be useless for selector generation.\",\n );\n }\n return html;\n}\n\nfunction looksLikeChallenge(html: string, title?: string): boolean {\n const t = (title ?? \"\").toLowerCase();\n if (\n t.includes(\"just a moment\") ||\n t.includes(\"attention required\") ||\n t.includes(\"access denied\") ||\n t.includes(\"cloudflare\") ||\n t.includes(\"ddos-guard\")\n ) return true;\n const head = html.slice(0, 4000).toLowerCase();\n return (\n head.includes(\"/cdn-cgi/challenge-platform\") ||\n head.includes(\"cf-chl-bypass\") ||\n head.includes(\"cf-browser-verification\") ||\n head.includes(\"checking if the site connection is secure\") ||\n head.includes(\"enable javascript and cookies to continue\") ||\n head.includes(\"__cf_chl_\") ||\n head.includes(\"perimeterx\") ||\n head.includes(\"px-captcha\") ||\n head.includes(\"hcaptcha-challenge\") ||\n head.includes(\"ddos protection by\")\n );\n}\n\nasync function safeText(res: Response): Promise {\n try { return await res.text(); } catch { return \"\"; }\n}\n","deleted","label",[113],"ad56301a-4d5d-11f1-8ee7-ee650bb23af1","main.ts","6bc5243a-4d5d-11f1-ab91-ee650bb23af1",2,["D",1778520016027],1,"http","// HTTP entrypoint: POST { url } -> Playwright .ts test as plain text.\n// Auth: X-Auth-Key header must equal env AUTH_KEY.\n\nimport { fetchHtml, CloudflareChallengeError, TinyFishError } from \"./tinyfish.ts\";\nimport { cleanHtml } from \"./clean.ts\";\nimport { generatePlaywrightTest, LlmError } from \"./llm.ts\";\n\nconst HARD_BUDGET_MS = 55_000; // stay under the 60s ceiling\n\nfunction json(status: number, body: Record) {\n return new Response(JSON.stringify(body), {\n status,\n headers: { \"content-type\": \"application/json\" },\n });\n}\n\nfunction text(status: number, body: string, contentType = \"text/plain; charset=utf-8\") {\n return new Response(body, { status, headers: { \"content-type\": contentType } });\n}\n\nexport default async function handler(req: Request): Promise {\n if (req.method === \"GET\") {\n return json(200, {\n ok: true,\n service: \"playwright-gen\",\n usage: \"POST { url } with header X-Auth-Key. Returns a Playwright .ts test file as text/plain.\",\n env_required: [\"AUTH_KEY\", \"TINYFISH_API_KEY\", \"LLM_TOKEN\", \"LLM_BASE_URL\", \"LLM_MODEL\"],\n });\n }\n if (req.method !== \"POST\") {\n return json(405, { error: \"method_not_allowed\" });\n }\n\n const authKey = Deno.env.get(\"AUTH_KEY\");\n if (!authKey) return json(500, { error: \"server_misconfigured\", detail: \"AUTH_KEY not set\" });\n if (req.headers.get(\"x-auth-key\") !== authKey) {\n return json(401, { error: \"unauthorized\" });\n }\n\n let body: { url?: unknown };\n try {\n body = await req.json();\n } catch {\n return json(400, { error: \"invalid_json\" });\n }\n const url = typeof body.url === \"string\" ? body.url.trim() : \"\";\n if (!url || !/^https?:\\/\\//i.test(url)) {\n return json(400, { error: \"invalid_url\", detail: \"Provide a JSON body { url: 'https://...' }\" });\n }\n\n const tinyfishKey = Deno.env.get(\"TINYFISH_API_KEY\");\n const llmToken = Deno.env.get(\"LLM_TOKEN\");\n const llmBaseUrl = Deno.env.get(\"LLM_BASE_URL\");\n const llmModel = Deno.env.get(\"LLM_MODEL\");\n if (!tinyfishKey) return json(500, { error: \"server_misconfigured\", detail: \"TINYFISH_API_KEY not set\" });\n if (!llmToken) return json(500, { error: \"server_misconfigured\", detail: \"LLM_TOKEN not set\" });\n if (!llmBaseUrl) return json(500, { error: \"server_misconfigured\", detail: \"LLM_BASE_URL not set (e.g. https://api.openai.com/v1)\" });\n if (!llmModel) return json(500, { error: \"server_misconfigured\", detail: \"LLM_MODEL not set\" });\n\n const deadline = Date.now() + HARD_BUDGET_MS;\n const remaining = () => Math.max(1000, deadline - Date.now());\n\n let rawHtml: string;\n try {\n rawHtml = await fetchHtml({\n url,\n apiKey: tinyfishKey,\n timeoutMs: Math.min(20_000, remaining()),\n });\n } catch (e) {\n if (e instanceof CloudflareChallengeError) {\n return json(502, { error: \"target_blocked_by_anti_bot\", detail: e.message });\n }\n if (e instanceof TinyFishError) {\n return json(502, { error: \"tinyfish_failed\", detail: e.message });\n }\n return json(500, { error: \"unexpected\", detail: String(e) });\n }\n\n const cleaned = cleanHtml(rawHtml);\n if (!cleaned || cleaned.length < 200) {\n return json(502, { error: \"empty_after_cleaning\", detail: \"Cleaned HTML was too small to be useful.\" });\n }\n\n let testFile: string;\n try {\n testFile = await generatePlaywrightTest({\n url,\n html: cleaned,\n baseUrl: llmBaseUrl,\n token: llmToken,\n model: llmModel,\n timeoutMs: remaining(),\n });\n } catch (e) {\n if (e instanceof LlmError) {\n return json(502, { error: \"llm_failed\", detail: e.message });\n }\n return json(500, { error: \"unexpected\", detail: String(e) });\n }\n\n return text(200, testFile);\n}\n","tinyfish",[130],"d411f9e6-4d5d-11f1-aee2-ee650bb23af1","llm.ts","d411fa90-4d5d-11f1-aee2-ee650bb23af1",5,["D",1778520081035],"// OpenAI-compatible chat completions client.\n// Returns the raw assistant text — which the system prompt instructs to be a\n// pure Playwright TypeScript test file.\n\nexport class LlmError extends Error {}\n\ninterface GenArgs {\n url: string;\n html: string;\n baseUrl: string; // e.g. https://api.openai.com/v1\n token: string;\n model: string;\n timeoutMs: number;\n}\n\nconst SYSTEM_PROMPT = [\n \"You generate Playwright TypeScript test files.\",\n \"Given a cleaned HTML snapshot of one page, identify the page's critical user journeys and the most important robust functional assertions, and produce a single .ts test file that exercises them.\",\n \"\",\n \"Locator priority (use strictly in this order, fall back only when the previous is impossible):\",\n \"1) page.getByRole(role, { name })\",\n \"2) page.getByLabel(text)\",\n \"3) page.getByText(text)\",\n \"4) page.getByPlaceholder(text)\",\n \"5) page.getByAltText(text)\",\n \"6) page.getByTitle(text)\",\n \"7) page.getByTestId(id) // only when data-testid / data-test / data-qa / data-cy exists\",\n \"8) page.locator(css) // last resort, css only, anchored on stable attributes\",\n \"\",\n \"Hard rules — never use:\",\n \"- nth-child / index-based selectors\",\n \"- full XPath or absolute paths\",\n \"- hashed / CSS-in-JS / BEM-style class names\",\n \"- deep parent-child chains (.a > .b > .c)\",\n \"- inline-style or computed-style selectors\",\n \"- :hover / :active or any browser-variant pseudo-classes\",\n \"- page.$ / page.$$ (use locators)\",\n \"- dynamic ARIA values like aria-controls='popup-123'\",\n \"\",\n \"Filtering & narrowing: prefer .filter({ hasText }), .filter({ has }), getByRole scoping.\",\n \"Always rely on Playwright's auto-waiting via expect() and locator actions. No manual sleeps.\",\n \"\",\n \"Output requirements:\",\n \"- Output ONLY the .ts file contents. No markdown fences, no commentary, no explanation.\",\n \"- First line must be: import { test, expect } from '@playwright/test';\",\n \"- Use one test.describe block named after the page, and multiple test() blocks — one per critical journey/assertion.\",\n \"- Use page.goto with the original URL.\",\n \"- Assertions must be robust and functional (visibility of critical landmarks, presence of CTAs by role+name, form field labels, semantic page structure). Avoid asserting on copy that is likely to be localized or A/B tested unless it's the only stable anchor.\",\n \"- If the page is clearly a login/auth/checkout/search/article/listing/landing page, tailor the assertions to that template's critical journey.\",\n].join(\"\\n\");\n\nexport async function generatePlaywrightTest(args: GenArgs): Promise {\n const { url, html, baseUrl, token, model, timeoutMs } = args;\n\n const endpoint = baseUrl.replace(/\\/+$/, \"\") + \"/chat/completions\";\n const userContent = [\n `URL: ${url}`,\n \"\",\n \"Cleaned HTML snapshot (body only, junk attributes/classes/scripts already stripped):\",\n \"```html\",\n html,\n \"```\",\n ].join(\"\\n\");\n\n const ctrl = new AbortController();\n const t = setTimeout(() => ctrl.abort(), timeoutMs);\n\n let res: Response;\n try {\n res = await fetch(endpoint, {\n method: \"POST\",\n headers: {\n \"content-type\": \"application/json\",\n authorization: `Bearer ${token}`,\n },\n body: JSON.stringify({\n model,\n temperature: 0.1,\n messages: [\n { role: \"system\", content: SYSTEM_PROMPT },\n { role: \"user\", content: userContent },\n ],\n }),\n signal: ctrl.signal,\n });\n } catch (e) {\n clearTimeout(t);\n if ((e as Error).name === \"AbortError\") {\n throw new LlmError(`LLM request timed out after ${timeoutMs}ms`);\n }\n throw new LlmError(`LLM request failed: ${String(e)}`);\n }\n clearTimeout(t);\n\n if (!res.ok) {\n const detail = await safeText(res);\n throw new LlmError(`LLM HTTP ${res.status}: ${detail.slice(0, 800)}`);\n }\n\n let payload: {\n choices?: { message?: { content?: string } }[];\n error?: { message?: string };\n };\n try {\n payload = await res.json();\n } catch {\n throw new LlmError(\"LLM returned non-JSON\");\n }\n if (payload.error) {\n throw new LlmError(`LLM error: ${payload.error.message ?? JSON.stringify(payload.error)}`);\n }\n const content = payload.choices?.[0]?.message?.content?.trim();\n if (!content) throw new LlmError(\"LLM returned empty content\");\n\n return stripCodeFences(content);\n}\n\nfunction stripCodeFences(s: string): string {\n // Defensive: if the model wraps output in ```ts ... ``` despite instructions.\n const fenced = s.match(/^```(?:ts|typescript)?\\s*([\\s\\S]*?)\\s*```$/);\n return fenced ? fenced[1].trim() : s;\n}\n\nasync function safeText(res: Response): Promise {\n try { return await res.text(); } catch { return \"\"; }\n}\n",[140],"6cf705ee-4d62-11f1-bedd-ee650bb23af1","clean.ts","c62f8046-4d5d-11f1-8b48-ee650bb23af1",["D",1778522055495],"// Aggressive HTML cleaner for LLM consumption.\n// Uses deno-dom (WASM, no native deps — works on Val Town's Deno runtime).\n// Goal: strip everything hostile to robust Playwright selectors and keep only\n// what drives getByRole / getByLabel / getByText / getByPlaceholder / getByAltText /\n// getByTitle / getByTestId, plus last-resort CSS.\n\nimport { DOMParser, Element, Node } from \"https://deno.land/x/deno_dom@v0.1.49/deno-dom-wasm.ts\";\n\nconst MAX_OUTPUT_BYTES = 120_000;\n\n// deno-dom Node type constants (not exposed as Node.ELEMENT_NODE etc.)\nconst ELEMENT_NODE = 1;\nconst TEXT_NODE = 3;\nconst COMMENT_NODE = 8;\n\nconst KILL_TAGS = new Set([\n \"script\", \"style\", \"noscript\", \"template\", \"link\", \"meta\",\n \"iframe\",\n \"svg\", \"canvas\", \"audio\", \"video\", \"source\", \"track\", \"object\", \"embed\", \"picture\",\n \"head\", \"base\",\n]);\n\nconst KEEP_ATTRS = new Set([\n \"id\", \"name\", \"type\", \"value\", \"href\", \"alt\", \"title\", \"placeholder\", \"for\",\n \"role\", \"aria-label\", \"aria-labelledby\", \"aria-describedby\",\n \"aria-current\", \"aria-expanded\", \"aria-pressed\", \"aria-checked\", \"aria-selected\",\n \"aria-disabled\", \"aria-required\", \"aria-invalid\", \"aria-haspopup\",\n \"label\", \"checked\", \"selected\", \"disabled\", \"readonly\", \"required\", \"multiple\",\n \"min\", \"max\", \"step\", \"pattern\", \"autocomplete\", \"inputmode\",\n \"contenteditable\",\n]);\n\nconst KEEP_DATA_ATTRS = new Set([\n \"data-testid\", \"data-test\", \"data-test-id\", \"data-qa\", \"data-cy\",\n]);\n\nconst HASHED_CLASS_RE = /^(css-|sc-|jsx-|emotion-|chakra-|mui-|MuiBox-|mantine-|tw-)/i;\nconst RANDOMISH_RE = /[a-z]_[a-zA-Z0-9]{4,}$|^[A-Za-z]+-[a-z0-9]{5,}$|^[a-f0-9]{6,}$/;\n\nfunction isJunkClass(c: string): boolean {\n if (!c) return true;\n if (HASHED_CLASS_RE.test(c)) return true;\n if (c.includes(\"__\") || c.includes(\"--\")) return true;\n if (RANDOMISH_RE.test(c)) return true;\n return false;\n}\n\nfunction cleanClassAttr(value: string): string {\n const kept = value\n .split(/\\s+/)\n .filter(Boolean)\n .filter((c) => !isJunkClass(c));\n return kept.slice(0, 4).join(\" \");\n}\n\nconst DROP_ATTR_PATTERNS: RegExp[] = [\n /^on[a-z]+/i,\n /^style$/i,\n /^xmlns/i,\n /^ng-/i, /^_ngcontent/i, /^_nghost/i,\n /^v-/i, /^data-v-[a-f0-9]+/i,\n /^data-reactid$/i, /^data-react/i,\n /^data-svelte-h$/i,\n /^data-astro/i,\n /^itemscope$/i, /^itemtype$/i, /^itemprop$/i,\n /^crossorigin$/i, /^integrity$/i, /^nonce$/i, /^referrerpolicy$/i,\n /^loading$/i, /^decoding$/i, /^fetchpriority$/i,\n /^srcset$/i, /^sizes$/i,\n /^translate$/i, /^spellcheck$/i, /^autocapitalize$/i, /^autocorrect$/i,\n /^enterkeyhint$/i,\n];\n\nfunction shouldDropAttr(name: string): boolean {\n const n = name.toLowerCase();\n if (KEEP_ATTRS.has(n)) return false;\n if (n.startsWith(\"data-\")) return !KEEP_DATA_ATTRS.has(n);\n return DROP_ATTR_PATTERNS.some((re) => re.test(n));\n}\n\nfunction isHiddenByInline(el: Element): boolean {\n if (el.hasAttribute(\"hidden\")) return true;\n const ariaHidden = el.getAttribute(\"aria-hidden\");\n if (ariaHidden === \"true\") return true;\n const style = el.getAttribute(\"style\") ?? \"\";\n if (/display\\s*:\\s*none/i.test(style)) return true;\n if (/visibility\\s*:\\s*hidden/i.test(style)) return true;\n return false;\n}\n\nconst PRUNE_IF_LARGE = [\"footer\", \"aside\", \"nav\"] as const;\n\nfunction walkClean(root: Element): void {\n // Snapshot the list first; mutating during traversal breaks deno-dom's NodeList.\n const all = Array.from(root.querySelectorAll(\"*\")) as Element[];\n const toRemove: Element[] = [];\n for (const node of all) {\n const tag = node.tagName?.toLowerCase();\n if (!tag) continue;\n if (KILL_TAGS.has(tag)) {\n toRemove.push(node);\n continue;\n }\n if (isHiddenByInline(node)) {\n toRemove.push(node);\n continue;\n }\n // Snapshot attributes too — removing during iteration is unsafe.\n const attrNames: string[] = [];\n for (const a of node.attributes) attrNames.push(a.name);\n for (const name of attrNames) {\n if (shouldDropAttr(name)) {\n node.removeAttribute(name);\n continue;\n }\n if (name === \"class\") {\n const cleaned = cleanClassAttr(node.getAttribute(\"class\") ?? \"\");\n if (cleaned) node.setAttribute(\"class\", cleaned);\n else node.removeAttribute(\"class\");\n continue;\n }\n if (name === \"href\") {\n const v = node.getAttribute(\"href\") ?? \"\";\n if (v.startsWith(\"javascript:\")) {\n node.removeAttribute(\"href\");\n } else if (v.length > 300) {\n node.setAttribute(\"href\", v.slice(0, 300));\n }\n }\n if (name === \"src\") {\n const v = node.getAttribute(\"src\") ?? \"\";\n if (v.startsWith(\"data:\")) node.removeAttribute(\"src\");\n }\n }\n }\n for (const n of toRemove) n.remove();\n}\n\nfunction removeComments(node: Node): void {\n const children = Array.from(node.childNodes) as Node[];\n for (const c of children) {\n if (c.nodeType === COMMENT_NODE) {\n (c as unknown as { remove: () => void }).remove();\n } else if (c.nodeType === ELEMENT_NODE) {\n removeComments(c);\n }\n }\n}\n\nfunction collapseWhitespace(root: Node): void {\n const children = Array.from(root.childNodes) as Node[];\n for (const c of children) {\n if (c.nodeType === TEXT_NODE) {\n const t = (c as unknown as { data: string }).data ?? \"\";\n (c as unknown as { data: string }).data = t.replace(/\\s+/g, \" \");\n } else if (c.nodeType === ELEMENT_NODE) {\n collapseWhitespace(c);\n }\n }\n}\n\nfunction depth(el: Element): number {\n let d = 0;\n let p: Element | null = el.parentElement;\n while (p) { d++; p = p.parentElement; }\n return d;\n}\n\nfunction hasUsefulAttrs(el: Element): boolean {\n for (const a of el.attributes) {\n const n = a.name.toLowerCase();\n if (KEEP_ATTRS.has(n)) return true;\n if (KEEP_DATA_ATTRS.has(n)) return true;\n if (n === \"class\" && (a.value ?? \"\").trim()) return true;\n }\n return false;\n}\n\nfunction bytes(s: string): number {\n return new TextEncoder().encode(s).length;\n}\n\nfunction progressivePrune(body: Element): void {\n for (const tag of PRUNE_IF_LARGE) {\n if (bytes(body.outerHTML) <= MAX_OUTPUT_BYTES) return;\n const nodes = Array.from(body.querySelectorAll(tag)) as Element[];\n for (const n of nodes) n.remove();\n }\n if (bytes(body.outerHTML) > MAX_OUTPUT_BYTES) {\n const divs = Array.from(body.querySelectorAll(\"div\")) as Element[];\n divs.sort((a, b) => depth(b) - depth(a));\n for (const d of divs) {\n if (bytes(body.outerHTML) <= MAX_OUTPUT_BYTES) return;\n const childCount = d.children?.length ?? 0;\n const text = (d.textContent ?? \"\").trim();\n if (!hasUsefulAttrs(d) && childCount === 0 && !text) {\n d.remove();\n }\n }\n }\n}\n\nexport function cleanHtml(input: string): string {\n const doc = new DOMParser().parseFromString(input, \"text/html\");\n if (!doc) return \"\";\n const body = (doc.querySelector(\"body\") ?? doc.documentElement) as Element | null;\n if (!body) return \"\";\n\n removeComments(body);\n walkClean(body);\n collapseWhitespace(body);\n\n let serialized = body.outerHTML;\n if (bytes(serialized) > MAX_OUTPUT_BYTES) {\n progressivePrune(body);\n serialized = body.outerHTML;\n }\n if (bytes(serialized) > MAX_OUTPUT_BYTES) {\n serialized = serialized.slice(0, MAX_OUTPUT_BYTES) + \"\\n\";\n }\n return serialized;\n}\n",[147],"0790e58e-4d5e-11f1-8c43-ee650bb23af1","README.md","f564f1de-4d5d-11f1-a562-ee650bb23af1",7,["D",1778520167424],"# playwright-gen\n\nPublic HTTP val that turns a URL into a Playwright `.ts` test file using TinyFish (real-browser HTML fetch) + an OpenAI-compatible LLM (selector & assertion generation).\n\n## Endpoint\n\n`POST https://khawjaahmad--6bc5243a4d5d11f1ab91ee650bb23af1.web.val.run`\n\nHeaders:\n- `X-Auth-Key: ` — required\n- `Content-Type: application/json`\n\nBody:\n```json\n{ \"url\": \"https://example.com/login\" }\n```\n\nResponse: `text/plain` containing a Playwright TypeScript test file. Save as `*.spec.ts` and run.\n\n`GET /` returns a small JSON usage blurb.\n\n## Environment variables\n\nSet these in the val's settings (already created with placeholders — replace before first use):\n\n| Key | Purpose |\n|--------------------|-------------------------------------------------------------------------|\n| `AUTH_KEY` | Shared secret gating the endpoint. Generate a long random string. |\n| `TINYFISH_API_KEY` | https://agent.tinyfish.ai/api-keys |\n| `LLM_TOKEN` | Bearer token for your OpenAI-compatible provider |\n| `LLM_BASE_URL` | Base URL ending in `/v1` (OpenAI, OpenRouter, Groq, Together, Gemini-OpenAI, ...) |\n| `LLM_MODEL` | Model name passed to `chat/completions` |\n\n## Why an auth key on a public val?\n\nThe val is public so its source is visible, but the source doesn't contain secrets — they live in env. The `AUTH_KEY` check prevents random callers from running up your TinyFish + LLM bills via your public URL. Without it, anyone who finds the endpoint can drain your accounts.\n\n## Pipeline\n\n1. Auth check (`X-Auth-Key`)\n2. TinyFish Fetch with `format: \"html\"` (timeout 20s)\n3. Aggressive cleaning (linkedom):\n - Drop `

Search