[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_212":213,"_214":215},"routes/_app.x.$handle.$name",{"_94":93,"_32":45,"_102":84,"_27":103,"_13":14,"_104":-5,"_11":105,"_106":107,"_108":109,"_110":111,"_112":113,"_20":-5,"_114":-5,"_115":-5,"_116":25,"_117":118,"_119":120,"_121":-5,"_122":-5,"_123":124},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_43":44,"_32":45,"_38":46,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":-5,"_57":25,"_58":-5,"_59":-5,"_60":61,"_62":63,"_64":-5,"_65":-5,"_66":-5,"_67":-5,"_68":69,"_70":69,"_71":72,"_73":72,"_74":72,"_75":69,"_76":69,"_77":78,"_79":80,"_81":82,"_83":25},"version",15,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":28,"_33":34,"_35":37,"_38":39,"_40":14,"_41":42,"_24":25},"previous",{"_30":31,"_32":28,"_33":34,"_35":36},"deleted",false,"deletedHere","path","main.ts","latestPath","fileId","f3c77a2a-3f3b-11f1-9116-42b51c65c3df","name","fileType","http","content","// ============================================================================\n// SEC Filings Signal Scraper — v4 (catch-all mode)\n// Philosophy: capture everything that might be signal. Don't filter.\n// Let the downstream LLM decide what matters.\n// ============================================================================\n\nconst HEADERS = {\n \"User-Agent\": \"AlexKrasiuk alex@myrunbook.com\",\n \"Accept-Encoding\": \"gzip, deflate\",\n};\n\nconst CONTEXT_CHARS = 800; // WIDER context — give LLM more to work with\nconst MAX_HITS_RETURNED = 150; // Cap payload size, not per-keyword\n\n// Broad keyword list — cast a wide net\nconst KEYWORDS = [\n // AI / ML\n \"artificial intelligence\",\n \"AI agent\",\n \"AI agents\",\n \"generative AI\",\n \"gen AI\",\n \"GenAI\",\n \"machine learning\",\n \"large language model\",\n \"LLM\",\n \"AI-enabled\",\n \"AI-powered\",\n \"AI initiative\",\n \"AI technology\",\n \"AI tool\",\n \"neural network\",\n \"deep learning\",\n\n // Headcount / workforce\n \"headcount reduction\",\n \"headcount reductions\",\n \"workforce reduction\",\n \"reduction in force\",\n \"reduction-in-force\",\n \"workforce optimization\",\n \"restructuring\",\n \"severance\",\n \"severance charges\",\n \"layoff\",\n \"layoffs\",\n \"job cut\",\n \"overhead reduction\",\n \"workforce transformation\",\n \"reorganization\",\n \"reorganize\",\n\n // Back office / ops\n \"back office\",\n \"back-office\",\n \"back office function\",\n \"shared services\",\n \"shared service\",\n \"shared service center\",\n \"administrative function\",\n \"administrative functions\",\n \"centralize\",\n \"centralization\",\n \"offshore\",\n \"offshoring\",\n \"outsourc\",\n\n // Automation\n \"process automation\",\n \"robotic process automation\",\n \"RPA\",\n \"automate\",\n \"automated\",\n \"automation\",\n \"AI-enabled automation\",\n \"intelligent automation\",\n \"workflow automation\",\n \"workflow tool\",\n\n // Transformation\n \"digital transformation\",\n \"technology transformation\",\n \"digital initiative\",\n \"digital strategy\",\n \"modernize\",\n \"modernization\",\n \"technology investment\",\n \"technology upgrade\",\n\n // Efficiency\n \"process optimization\",\n \"process improvement\",\n \"operational efficiency\",\n \"operational excellence\",\n \"productivity improvement\",\n \"productivity gain\",\n \"cost savings\",\n \"cost reduction\",\n \"cost optimization\",\n \"margin expansion\",\n \"margin improvement\",\n \"do more with less\",\n \"scale without\",\n \"efficiency initiative\",\n \"efficiency program\",\n\n // Manual work\n \"manual processes\",\n \"manual work\",\n \"manual task\",\n \"streamline\",\n \"streamlined\",\n \"streamlining\",\n\n // Field ops / dispatch-adjacent (Runbook-specific)\n \"dispatch\",\n \"scheduling\",\n \"route optimization\",\n \"field service\",\n \"field operations\",\n \"field technician\",\n \"service technician\",\n \"work order\",\n \"customer service center\",\n \"call center\",\n \"contact center\",\n\n // Financial pressure tells\n \"margin pressure\",\n \"inflation\",\n \"wage inflation\",\n \"labor cost\",\n \"labor shortage\",\n];\n\n// ============================================================================\n// Ticker cache (24h)\n// ============================================================================\n\nlet tickerCache: { data: any; ts: number } | null = null;\nconst CACHE_TTL_MS = 24 * 60 * 60 * 1000;\n\nasync function getTickers(): Promise {\n if (tickerCache && Date.now() - tickerCache.ts < CACHE_TTL_MS) {\n return tickerCache.data;\n }\n const data = JSON.parse(\n await secFetch(\"https://www.sec.gov/files/company_tickers.json\"),\n );\n tickerCache = { data, ts: Date.now() };\n return data;\n}\n\n// ============================================================================\n// SEC fetch with 429 retry\n// ============================================================================\n\nasync function secFetch(url: string, retries = 3): Promise {\n for (let i = 0; i < retries; i++) {\n const resp = await fetch(url, { headers: HEADERS });\n if (resp.ok) return resp.text();\n if (resp.status === 429) {\n await new Promise((r) => setTimeout(r, 1000 * (i + 1)));\n continue;\n }\n throw new Error(`SEC fetch failed: ${resp.status} for ${url}`);\n }\n throw new Error(`SEC fetch failed after ${retries} retries for ${url}`);\n}\n\n// ============================================================================\n// HTML stripping\n// ============================================================================\n\nfunction stripHtml(html: string): string {\n let text = html.replace(/]*>.*?<\\/script>/gis, \" \");\n text = text.replace(/]*>.*?<\\/style>/gis, \" \");\n text = text.replace(/<\\/?ix:[^>]+>/gi, \" \");\n text = text.replace(/<[^>]+>/g, \" \");\n text = text\n .replace(/&/g, \"&\").replace(/</g, \"<\").replace(/>/g, \">\")\n .replace(/"/g, '\"').replace(/'/g, \"'\").replace(/ /g, \" \")\n .replace(/&#\\d+;/g, \" \").replace(/&[a-zA-Z]+;/g, \" \");\n text = text.replace(/\\s+/g, \" \").trim();\n return text;\n}\n\n// ============================================================================\n// Analysis helpers\n// ============================================================================\n\nfunction normalize(s: string): string {\n return s.toLowerCase().replace(/[^a-z0-9\\s]/g, \"\").replace(/\\s+/g, \" \")\n .trim();\n}\n\nfunction findSection(text: string, matchIdx: number): string {\n const before = text.slice(Math.max(0, matchIdx - 10000), matchIdx);\n const sectionMatches = [\n ...before.matchAll(\n /(?:^|\\.\\s+|\\n\\s*)Item\\s+(\\d+[A-Z]?)\\.\\s+([A-Z][A-Za-z\\s&',\\-]{3,80}?)(?=\\s*(?:\\.|\\n|Table of Contents))/g,\n ),\n ];\n if (sectionMatches.length === 0) return \"Unknown\";\n const last = sectionMatches[sectionMatches.length - 1];\n const title = last[2].trim().replace(/\\s+/g, \" \").replace(/\\s+of the$/i, \"\");\n return `Item ${last[1]} - ${title}`;\n}\n\n// Score is INFORMATIONAL ONLY — never used to filter\n// Gives the downstream LLM a hint about which passages are likely richest\nfunction scoreHit(context: string, section: string, keyword: string): number {\n const lower = context.toLowerCase();\n let score = 0;\n\n // Strategic neighbor words\n const neighbors = [\n \"initiative\",\n \"invest\",\n \"deploy\",\n \"implement\",\n \"launch\",\n \"reduce\",\n \"reduction\",\n \"eliminate\",\n \"transform\",\n \"strategy\",\n \"pilot\",\n \"acquired\",\n \"acquisition\",\n \"partnership\",\n \"committed\",\n \"priority\",\n \"focus\",\n \"accelerate\",\n \"expand\",\n ];\n for (const w of neighbors) if (lower.includes(w)) score += 1;\n\n // Section hints\n const s = section.toLowerCase();\n if (s.startsWith(\"item 7 \")) score += 3;\n if (s.startsWith(\"item 1a\")) score += 2;\n if (s.startsWith(\"item 2 \") && s.includes(\"management\")) score += 2;\n\n // Mega-signal patterns\n const hasStreamline = /\\bstreamlin/i.test(lower);\n const hasHeadcount =\n /\\b(headcount|severance|workforce reduction|reduction in force|layoff)/i\n .test(lower);\n const hasDollar = /\\$\\s*\\d+(?:\\.\\d+)?\\s*(?:million|billion|M|B)\\b/i.test(\n context,\n );\n const hasBackOffice =\n /\\b(back[\\s-]?office|administrative function|shared service)/i.test(lower);\n\n if (hasStreamline && hasHeadcount) score += 8;\n if (hasStreamline && hasDollar) score += 4;\n if (hasHeadcount && hasDollar) score += 4;\n if (hasBackOffice) score += 3;\n\n return score;\n}\n\n// EXACT dedupe only — remove only bit-for-bit duplicates\n// Preserves near-duplicates because they may contain slightly different useful context\nfunction buildDedupeKey(context: string): string {\n return normalize(context);\n}\n\nfunction countByKeyword(hits: { keyword: string }[]): Record {\n const counts: Record = {};\n for (const h of hits) counts[h.keyword] = (counts[h.keyword] || 0) + 1;\n return counts;\n}\n\n// ============================================================================\n// Main handler\n// ============================================================================\n\nexport default async function handler(req: Request): Promise {\n const url = new URL(req.url);\n let ticker = (url.searchParams.get(\"ticker\") || \"\").toUpperCase().trim();\n\n if (!ticker && req.method === \"POST\") {\n try {\n const body = await req.json();\n ticker = (body.ticker || \"\").toUpperCase().trim();\n } catch { /* ignore */ }\n }\n\n if (!ticker) {\n return Response.json(\n {\n error: 'Pass ?ticker=AAPL or POST with {\"ticker\": \"AAPL\"}',\n usage: {\n get: \"GET /?ticker=AAPL\",\n post: 'POST / with {\"ticker\": \"AAPL\"}',\n },\n },\n { status: 400 },\n );\n }\n\n try {\n // Step 1: Ticker -> CIK\n const tickersJson = await getTickers();\n let cik: string | null = null;\n let company: string | null = null;\n for (const entry of Object.values(tickersJson) as any[]) {\n if (entry.ticker.toUpperCase() === ticker) {\n cik = String(entry.cik_str).padStart(10, \"0\");\n company = entry.title;\n break;\n }\n }\n if (!cik) {\n return Response.json(\n { error: `Ticker ${ticker} not found in SEC EDGAR` },\n { status: 404 },\n );\n }\n\n // Step 2: Latest 10-K and 10-Q\n const subs = JSON.parse(\n await secFetch(`https://data.sec.gov/submissions/CIK${cik}.json`),\n );\n const recent = subs.filings.recent;\n\n type FilingInfo = { url: string; date: string; type: string };\n const filings: FilingInfo[] = [];\n\n let found10K = false;\n let found10Q = false;\n for (let i = 0; i < recent.form.length; i++) {\n const form = recent.form[i];\n if ((form === \"10-K\" && !found10K) || (form === \"10-Q\" && !found10Q)) {\n const acc = recent.accessionNumber[i].replace(/-/g, \"\");\n const doc = recent.primaryDocument[i];\n filings.push({\n url: `https://www.sec.gov/Archives/edgar/data/${\n parseInt(cik)\n }/${acc}/${doc}`,\n date: recent.filingDate[i],\n type: form,\n });\n if (form === \"10-K\") found10K = true;\n if (form === \"10-Q\") found10Q = true;\n if (found10K && found10Q) break;\n }\n }\n\n if (filings.length === 0) {\n return Response.json({ error: `No 10-K or 10-Q found for ${ticker}` }, {\n status: 404,\n });\n }\n\n // Step 3: Scan each filing\n type Hit = {\n keyword: string;\n context: string;\n section: string;\n score: number;\n filing_type: string;\n filing_date: string;\n };\n const allHits: Hit[] = [];\n\n for (const filing of filings) {\n const html = await secFetch(filing.url);\n const text = stripHtml(html);\n\n for (const kw of KEYWORDS) {\n const pattern = new RegExp(\n \"\\\\b\" + kw.replace(/[.*+?^${}()|[\\]\\\\]/g, \"\\\\$&\") + \"\\\\b\",\n \"gi\",\n );\n let match;\n while ((match = pattern.exec(text)) !== null) {\n const charStart = Math.max(0, match.index - CONTEXT_CHARS);\n const charEnd = Math.min(\n text.length,\n match.index + kw.length + CONTEXT_CHARS,\n );\n const context = text.slice(charStart, charEnd).trim();\n const section = findSection(text, match.index);\n const score = scoreHit(context, section, kw);\n\n allHits.push({\n keyword: kw,\n context,\n section,\n score,\n filing_type: filing.type,\n filing_date: filing.date,\n });\n }\n }\n }\n\n // EXACT dedupe only (bit-for-bit identical passages)\n const seen = new Set();\n const unique: Hit[] = [];\n for (const h of allHits) {\n const key = buildDedupeKey(h.context);\n if (!seen.has(key)) {\n seen.add(key);\n unique.push(h);\n }\n }\n\n // Sort by score so LLM sees best stuff first — but DON'T filter\n unique.sort((a, b) => b.score - a.score);\n\n // Only cap is total payload size, not per-keyword\n const returned = unique.slice(0, MAX_HITS_RETURNED);\n\n return Response.json({\n ticker,\n company,\n filings: filings.map((f) => ({ type: f.type, date: f.date, url: f.url })),\n total_hits_found: unique.length,\n total_hits_returned: returned.length,\n keyword_counts: countByKeyword(unique),\n hits: returned, // flat list, sorted by score, no structure games\n });\n } catch (e: any) {\n return Response.json({ error: e.message }, { status: 500 });\n }\n}","// ============================================================================\n// SEC 10-K/10-Q Signal Scraper — v5\n// Changes from v4:\n// 1. FIXED section parser — was using \\n\\s*\\ anchors that never match after\n// stripHtml() collapses whitespace. Now finds \"Item N.\" markers directly.\n// 2. FIXED dedupe — was comparing full normalized contexts, but overlapping\n// context windows produce different strings. Now dedupes by keyword +\n// match position (within 400 chars = same passage).\n// 3. NEW 200KB hard cap — graceful trim if response exceeds Clay's limit.\n// 4. NEW note-section detection (e.g., \"Note 5 - Restructuring\").\n// ============================================================================\n\nconst HEADERS = {\n \"User-Agent\": \"AlexKrasiuk alex@myrunbook.com\",\n \"Accept-Encoding\": \"gzip, deflate\",\n};\n\nconst CONTEXT_CHARS = 800;\nconst MAX_HITS_RETURNED = 150;\nconst MAX_BYTES = 180_000; // Clay cell limit ~200KB, leave headroom\nconst DEDUPE_WINDOW = 400; // same-keyword hits within this distance = dupe\n\nconst KEYWORDS = [\n // AI / ML\n \"artificial intelligence\",\n \"AI agent\",\n \"AI agents\",\n \"generative AI\",\n \"gen AI\",\n \"GenAI\",\n \"machine learning\",\n \"large language model\",\n \"LLM\",\n \"AI-enabled\",\n \"AI-powered\",\n \"AI initiative\",\n \"AI technology\",\n \"AI tool\",\n \"neural network\",\n \"deep learning\",\n // Headcount / workforce\n \"headcount reduction\",\n \"headcount reductions\",\n \"workforce reduction\",\n \"reduction in force\",\n \"reduction-in-force\",\n \"workforce optimization\",\n \"restructuring\",\n \"severance\",\n \"severance charges\",\n \"layoff\",\n \"layoffs\",\n \"job cut\",\n \"overhead reduction\",\n \"workforce transformation\",\n \"reorganization\",\n \"reorganize\",\n // Back office / ops\n \"back office\",\n \"back-office\",\n \"shared services\",\n \"shared service\",\n \"shared service center\",\n \"administrative function\",\n \"administrative functions\",\n \"centralize\",\n \"centralization\",\n \"offshore\",\n \"offshoring\",\n \"outsourc\",\n // Automation\n \"process automation\",\n \"robotic process automation\",\n \"RPA\",\n \"automate\",\n \"automated\",\n \"automation\",\n \"intelligent automation\",\n \"workflow automation\",\n // Transformation\n \"digital transformation\",\n \"technology transformation\",\n \"digital initiative\",\n \"digital strategy\",\n \"modernize\",\n \"modernization\",\n \"technology investment\",\n \"technology upgrade\",\n // Efficiency\n \"process optimization\",\n \"process improvement\",\n \"operational efficiency\",\n \"operational excellence\",\n \"productivity improvement\",\n \"productivity gain\",\n \"cost savings\",\n \"cost reduction\",\n \"cost optimization\",\n \"margin expansion\",\n \"margin improvement\",\n \"efficiency initiative\",\n \"efficiency program\",\n // Manual work\n \"manual processes\",\n \"manual work\",\n \"manual task\",\n \"streamline\",\n \"streamlined\",\n \"streamlining\",\n // Field ops\n \"dispatch\",\n \"scheduling\",\n \"route optimization\",\n \"field service\",\n \"field operations\",\n \"field technician\",\n \"service technician\",\n \"work order\",\n \"call center\",\n \"contact center\",\n // Financial pressure\n \"margin pressure\",\n \"labor cost\",\n \"labor shortage\",\n];\n\n// Known 10-K item titles — used to produce clean section labels even when\n// the doc has weird formatting.\nconst ITEM_TITLES: Record = {\n \"1\": \"Business\",\n \"1A\": \"Risk Factors\",\n \"1B\": \"Unresolved Staff Comments\",\n \"1C\": \"Cybersecurity\",\n \"2\": \"Properties\",\n \"3\": \"Legal Proceedings\",\n \"4\": \"Mine Safety Disclosures\",\n \"5\": \"Market for Registrant's Common Equity\",\n \"6\": \"Selected Financial Data\",\n \"7\": \"Management's Discussion and Analysis\",\n \"7A\": \"Quantitative and Qualitative Disclosures About Market Risk\",\n \"8\": \"Financial Statements and Supplementary Data\",\n \"9\": \"Changes in and Disagreements with Accountants\",\n \"9A\": \"Controls and Procedures\",\n \"9B\": \"Other Information\",\n \"10\": \"Directors and Executive Officers\",\n \"11\": \"Executive Compensation\",\n \"12\": \"Security Ownership\",\n \"13\": \"Certain Relationships\",\n \"14\": \"Principal Accountant Fees\",\n \"15\": \"Exhibits\",\n};\n\n// ============================================================================\n// Ticker cache + fetch\n// ============================================================================\n\nlet tickerCache: { data: any; ts: number } | null = null;\nconst CACHE_TTL_MS = 24 * 60 * 60 * 1000;\n\nasync function secFetch(url: string, retries = 3): Promise {\n for (let i = 0; i < retries; i++) {\n const resp = await fetch(url, { headers: HEADERS });\n if (resp.ok) return resp.text();\n if (resp.status === 429) {\n await new Promise((r) => setTimeout(r, 1000 * (i + 1)));\n continue;\n }\n throw new Error(`SEC fetch failed: ${resp.status} for ${url}`);\n }\n throw new Error(`SEC fetch failed after ${retries} retries for ${url}`);\n}\n\nasync function getTickers(): Promise {\n if (tickerCache && Date.now() - tickerCache.ts < CACHE_TTL_MS) {\n return tickerCache.data;\n }\n const data = JSON.parse(\n await secFetch(\"https://www.sec.gov/files/company_tickers.json\"),\n );\n tickerCache = { data, ts: Date.now() };\n return data;\n}\n\nfunction stripHtml(html: string): string {\n let text = html.replace(/]*>.*?<\\/script>/gis, \" \");\n text = text.replace(/]*>.*?<\\/style>/gis, \" \");\n text = text.replace(/<\\/?ix:[^>]+>/gi, \" \");\n text = text.replace(/<[^>]+>/g, \" \");\n text = text\n .replace(/&/g, \"&\").replace(/</g, \"<\").replace(/>/g, \">\")\n .replace(/"/g, '\"').replace(/'/g, \"'\").replace(/ /g, \" \")\n .replace(/&#\\d+;/g, \" \").replace(/&[a-zA-Z]+;/g, \" \");\n text = text.replace(/\\s+/g, \" \").trim();\n return text;\n}\n\n// ============================================================================\n// FIXED Section detection\n// Strategy: find all \"Item N.\" markers in the text preceding the match.\n// The LAST one before matchIdx is the current section.\n// Also detects \"Note N.\" style markers for financial statement footnotes.\n// ============================================================================\n\ntype SectionMarker = {\n type: \"item\" | \"note\";\n code: string;\n title: string;\n idx: number;\n};\n\n// Pre-compute all section markers in a document once, then binary-search.\n// This is ~100x faster than re-scanning for each keyword hit.\n//\n// Design choice: rather than capture item titles from the text (fragile —\n// 10-Ks format them inconsistently), we match just the \"Item N.\" code and\n// look up the canonical title in ITEM_TITLES. Unknown codes are ignored so\n// we don't produce garbage like \"Item 99.1 - Press Release\".\nfunction indexSections(text: string): SectionMarker[] {\n const markers: SectionMarker[] = [];\n let m;\n\n // Item markers: \"Item 1.\", \"Item 1A.\", \"Item 7A.\" etc.\n // Require capital letter after the period so we don't match fractional\n // numbers or item cross-refs like \"Item 10.5a\" in exhibit lists.\n const itemRe = /\\bItem\\s+(\\d{1,2}[A-Z]?)\\.\\s+(?=[A-Z])/g;\n while ((m = itemRe.exec(text)) !== null) {\n const code = m[1];\n const title = ITEM_TITLES[code];\n if (!title) continue; // skip unknown codes\n markers.push({ type: \"item\", code, title, idx: m.index });\n }\n\n // Note markers in financial statements: \"5. RESTRUCTURING AND RELATED COSTS\".\n // Must be at sentence boundary, single/double digit, followed by ALL-CAPS title.\n const noteRe =\n /(?:\\.\\s+|\\s{2,})(\\d{1,2})\\s*\\.\\s+([A-Z][A-Z\\s&',\\-]{8,80}?)(?=\\s+[A-Z][a-z]|\\s+\\()/g;\n while ((m = noteRe.exec(text)) !== null) {\n const num = parseInt(m[1]);\n if (num < 1 || num > 25) continue;\n const title = m[2].trim().replace(/\\s+/g, \" \");\n if (title.length < 8) continue;\n markers.push({ type: \"note\", code: m[1], title, idx: m.index });\n }\n\n markers.sort((a, b) => a.idx - b.idx);\n return markers;\n}\n\nfunction findSection(markers: SectionMarker[], matchIdx: number): string {\n // Binary search for largest idx <= matchIdx\n if (markers.length === 0) return \"Unknown\";\n let lo = 0, hi = markers.length - 1, best = -1;\n while (lo <= hi) {\n const mid = (lo + hi) >> 1;\n if (markers[mid].idx <= matchIdx) {\n best = mid;\n lo = mid + 1;\n } else {\n hi = mid - 1;\n }\n }\n if (best < 0) return \"Unknown\";\n const m = markers[best];\n return m.type === \"item\"\n ? `Item ${m.code} - ${m.title}`\n : `Note ${m.code} - ${m.title}`;\n}\n\n// ============================================================================\n// Scoring (unchanged — informational, doesn't filter)\n// ============================================================================\n\nfunction scoreHit(context: string, section: string, _keyword: string): number {\n const lower = context.toLowerCase();\n let score = 0;\n\n const neighbors = [\n \"initiative\",\n \"invest\",\n \"deploy\",\n \"implement\",\n \"launch\",\n \"reduce\",\n \"reduction\",\n \"eliminate\",\n \"transform\",\n \"strategy\",\n \"pilot\",\n \"acquired\",\n \"acquisition\",\n \"partnership\",\n \"committed\",\n \"priority\",\n \"focus\",\n \"accelerate\",\n \"expand\",\n ];\n for (const w of neighbors) if (lower.includes(w)) score += 1;\n\n const s = section.toLowerCase();\n if (s.startsWith(\"item 7 \") || s.startsWith(\"item 7a\")) score += 3;\n if (s.startsWith(\"item 1a\")) score += 2;\n if (s.startsWith(\"note \") && /restructur|severance|impair/.test(s)) {\n score += 4;\n }\n\n const hasStreamline = /\\bstreamlin/i.test(lower);\n const hasHeadcount =\n /\\b(headcount|severance|workforce reduction|reduction in force|layoff)/i\n .test(lower);\n const hasDollar = /\\$\\s*\\d+(?:\\.\\d+)?\\s*(?:million|billion|M|B)\\b/i.test(\n context,\n );\n const hasBackOffice =\n /\\b(back[\\s-]?office|administrative function|shared service)/i.test(lower);\n\n if (hasStreamline && hasHeadcount) score += 8;\n if (hasStreamline && hasDollar) score += 4;\n if (hasHeadcount && hasDollar) score += 4;\n if (hasBackOffice) score += 3;\n\n return score;\n}\n\n// ============================================================================\n// FIXED position-based dedupe\n// Same keyword + match index within 400 chars = duplicate.\n// Keeps highest-scoring instance.\n// Different keywords at the same position are NOT deduped (valuable signal).\n// ============================================================================\n\ntype Hit = {\n keyword: string;\n context: string;\n section: string;\n score: number;\n filing_type: string;\n filing_date: string;\n _match_idx: number; // internal, stripped before return\n};\n\nfunction dedupeByPosition(hits: Hit[]): Hit[] {\n // Sort by score desc so highest-scoring wins when we collide\n const sorted = [...hits].sort((a, b) => b.score - a.score);\n // Track kept positions per (filing_date, keyword)\n const kept: Record = {};\n const out: Hit[] = [];\n for (const h of sorted) {\n const key = `${h.filing_date}|${h.keyword}`;\n const positions = kept[key] || [];\n const dupe = positions.some((p) =>\n Math.abs(p - h._match_idx) < DEDUPE_WINDOW\n );\n if (!dupe) {\n positions.push(h._match_idx);\n kept[key] = positions;\n out.push(h);\n }\n }\n // Re-sort by score for presentation\n return out.sort((a, b) => b.score - a.score);\n}\n\n// ============================================================================\n// Size cap — graceful trim to fit Clay's 200KB cell limit\n// ============================================================================\n\nfunction fitResponse(\n resp: T,\n): T {\n let json = JSON.stringify(resp);\n // Level 1: drop lowest-scored hits until under budget\n while (json.length > MAX_BYTES && resp.hits.length > 10) {\n resp.hits.pop();\n json = JSON.stringify(resp);\n }\n resp.total_hits_returned = resp.hits.length;\n // Level 2: shrink context windows\n if (json.length > MAX_BYTES) {\n for (const h of resp.hits) {\n if (h.context && h.context.length > 500) {\n h.context = h.context.slice(0, 500) + \"…\";\n }\n }\n }\n return resp;\n}\n\n// ============================================================================\n// Main handler\n// ============================================================================\n\nexport default async function handler(req: Request): Promise {\n const url = new URL(req.url);\n let ticker = (url.searchParams.get(\"ticker\") || \"\").toUpperCase().trim();\n\n if (!ticker && req.method === \"POST\") {\n try {\n const body = await req.json();\n ticker = (body.ticker || \"\").toUpperCase().trim();\n } catch { /* ignore */ }\n }\n\n if (!ticker) {\n return Response.json(\n { error: 'Pass ?ticker=AAPL or POST {\"ticker\": \"AAPL\"}' },\n { status: 400 },\n );\n }\n\n try {\n const tickersJson = await getTickers();\n let cik: string | null = null;\n let company: string | null = null;\n for (const entry of Object.values(tickersJson) as any[]) {\n if (entry.ticker.toUpperCase() === ticker) {\n cik = String(entry.cik_str).padStart(10, \"0\");\n company = entry.title;\n break;\n }\n }\n if (!cik) {\n return Response.json(\n { error: `Ticker ${ticker} not found in SEC EDGAR` },\n { status: 404 },\n );\n }\n\n const subs = JSON.parse(\n await secFetch(`https://data.sec.gov/submissions/CIK${cik}.json`),\n );\n const recent = subs.filings.recent;\n\n type FilingInfo = { url: string; date: string; type: string };\n const filings: FilingInfo[] = [];\n let found10K = false, found10Q = false;\n for (let i = 0; i < recent.form.length; i++) {\n const form = recent.form[i];\n if ((form === \"10-K\" && !found10K) || (form === \"10-Q\" && !found10Q)) {\n const acc = recent.accessionNumber[i].replace(/-/g, \"\");\n const doc = recent.primaryDocument[i];\n filings.push({\n url: `https://www.sec.gov/Archives/edgar/data/${\n parseInt(cik)\n }/${acc}/${doc}`,\n date: recent.filingDate[i],\n type: form,\n });\n if (form === \"10-K\") found10K = true;\n if (form === \"10-Q\") found10Q = true;\n if (found10K && found10Q) break;\n }\n }\n\n if (filings.length === 0) {\n return Response.json({ error: `No 10-K or 10-Q found for ${ticker}` }, {\n status: 404,\n });\n }\n\n const allHits: Hit[] = [];\n\n for (const filing of filings) {\n const html = await secFetch(filing.url);\n const text = stripHtml(html);\n const sectionMarkers = indexSections(text);\n\n for (const kw of KEYWORDS) {\n const pattern = new RegExp(\n \"\\\\b\" + kw.replace(/[.*+?^${}()|[\\]\\\\]/g, \"\\\\$&\") + \"\\\\b\",\n \"gi\",\n );\n let match;\n while ((match = pattern.exec(text)) !== null) {\n const charStart = Math.max(0, match.index - CONTEXT_CHARS);\n const charEnd = Math.min(\n text.length,\n match.index + kw.length + CONTEXT_CHARS,\n );\n const context = text.slice(charStart, charEnd).trim();\n const section = findSection(sectionMarkers, match.index);\n const score = scoreHit(context, section, kw);\n\n allHits.push({\n keyword: kw,\n context,\n section,\n score,\n filing_type: filing.type,\n filing_date: filing.date,\n _match_idx: match.index,\n });\n }\n }\n }\n\n const unique = dedupeByPosition(allHits);\n const returned = unique.slice(0, MAX_HITS_RETURNED).map((h) => {\n const { _match_idx, ...clean } = h;\n return clean;\n });\n\n const kwCounts: Record = {};\n for (const h of unique) {\n kwCounts[h.keyword] = (kwCounts[h.keyword] || 0) + 1;\n }\n const sectionCounts: Record = {};\n for (const h of unique) {\n sectionCounts[h.section] = (sectionCounts[h.section] || 0) + 1;\n }\n\n const response = {\n ticker,\n company,\n filings: filings.map((f) => ({ type: f.type, date: f.date, url: f.url })),\n total_hits_found: unique.length,\n total_hits_returned: returned.length,\n keyword_counts: kwCounts,\n section_counts: sectionCounts,\n hits: returned,\n };\n\n const trimmed = fitResponse(response);\n const finalJson = JSON.stringify(trimmed);\n return new Response(\n JSON.stringify({ ...trimmed, _response_size_bytes: finalJson.length }),\n { headers: { \"content-type\": \"application/json\" } },\n );\n } catch (e: any) {\n return Response.json({ error: e.message }, { status: 500 });\n }\n}","createdAt",["D",1776979764369],"branchVersion","branchId","f3c42c62-3f3b-11f1-9116-42b51c65c3df","id","f3c38384-3f3b-11f1-9116-42b51c65c3df","sec-10k-keyword-scraper",["D",1776966215049],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Scrape SEC EDGAR 10-K filings for AI/automation keywords","imageUrl","828451e7-38d3-43a6-9686-9dd07ed10f76","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","019dbb6e-9b96-734d-94c0-10d6676fe411","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_94":93,"_101":97},"org",{"_94":93,"_95":-5,"_96":97,"_18":55,"_98":99,"_100":25},"branch",{"_13":14,"_32":84,"_43":42,"_85":44,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":45,"_92":93},"isOwned","main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","alexkrasiuk","handle","displayName","avatar","https://avatars.githubusercontent.com/u/274322468?v=4","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_43":44,"_32":45,"_38":208,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":53,"_54":-5,"_18":55,"_56":-5,"_57":25,"_58":-5,"_59":-5,"_60":61,"_62":63,"_64":-5,"_65":-5,"_66":-5,"_67":-5,"_68":69,"_70":69,"_71":72,"_73":72,"_74":72,"_75":69,"_76":69,"_77":209,"_79":210,"_81":211,"_83":25},"tree",[183,184,185,186,187,188],"files",[183,184,185,186,187,188],"counts",{"_172":61,"_173":174,"_175":61,"_176":177,"_178":179},"contents",[125,126,127,128,129,130],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles",true,"grants","orgs","customDomains",{},{"_43":166,"_32":167,"_30":168,"_134":-5,"_41":42,"_40":150,"_136":120,"_38":169,"_13":61,"_33":138,"_35":170,"_24":25,"_140":-5,"_27":171},{"_43":159,"_32":160,"_30":161,"_134":-5,"_41":42,"_40":162,"_136":120,"_38":163,"_13":61,"_33":138,"_35":164,"_24":25,"_140":-5,"_27":165},{"_43":152,"_32":153,"_30":154,"_134":-5,"_41":42,"_40":155,"_136":120,"_38":156,"_13":61,"_33":138,"_35":157,"_24":25,"_140":-5,"_27":158},{"_43":148,"_32":28,"_30":31,"_134":-5,"_41":42,"_40":14,"_136":120,"_38":149,"_13":150,"_33":34,"_35":37,"_24":25,"_140":-5,"_27":151},{"_43":142,"_32":143,"_30":144,"_134":-5,"_41":42,"_40":145,"_136":120,"_38":146,"_13":61,"_33":138,"_35":103,"_24":25,"_140":-5,"_27":147},{"_43":131,"_32":132,"_30":133,"_134":-5,"_41":42,"_40":135,"_136":120,"_38":137,"_13":61,"_33":138,"_35":139,"_24":25,"_140":-5,"_27":141},"697512f4-3f47-11f1-9a9b-42b51c65c3df","sec-client.ts","697513b2-3f47-11f1-9a9b-42b51c65c3df","parentId",8,"latest",["D",1776971136950],"script","const HEADERS = {\n \"User-Agent\": \"AlexKrasiuk alex@myrunbook.com\",\n \"Accept-Encoding\": \"gzip, deflate\",\n};\n\n// Cache ticker->CIK lookup (refreshes every 24h)\nlet tickerCache: { data: any; ts: number } | null = null;\nconst CACHE_TTL_MS = 24 * 60 * 60 * 1000;\n\nexport async function getTickers(): Promise {\n if (tickerCache && Date.now() - tickerCache.ts < CACHE_TTL_MS) {\n return tickerCache.data;\n }\n const data = JSON.parse(await secFetch(\"https://www.sec.gov/files/company_tickers.json\"));\n tickerCache = { data, ts: Date.now() };\n return data;\n}\n\n// Fetch with retry + backoff for SEC rate limits (10 req/s)\nexport async function secFetch(url: string, retries = 3): Promise {\n for (let i = 0; i < retries; i++) {\n const resp = await fetch(url, { headers: HEADERS });\n if (resp.ok) return resp.text();\n if (resp.status === 429) {\n await new Promise(r => setTimeout(r, 1000 * (i + 1)));\n continue;\n }\n throw new Error(`SEC fetch failed: ${resp.status} ${resp.statusText} for ${url}`);\n }\n throw new Error(`SEC fetch failed after ${retries} retries for ${url}`);\n}\n","label",[132],"26b7874a-3f50-11f1-84b6-42b51c65c3df","new-file-8025.ts","26b78830-3f50-11f1-84b6-42b51c65c3df",14,["D",1776974890453],[143],"7fcca88c-3f5b-11f1-a85a-42b51c65c3df",["D",1776979764369],9,[28],"61195cdc-3f47-11f1-a505-42b51c65c3df","keywords.ts","61195db8-3f47-11f1-a505-42b51c65c3df",6,["D",1776971122935],"export const KEYWORDS = [\n // AI / ML\n \"artificial intelligence\", \"AI agent\", \"AI agents\",\n \"artificial intelligence initiative\",\n \"generative AI\", \"gen AI\", \"GenAI\",\n \"machine learning\", \"large language model\", \"LLM\",\n \"AI-enabled\", \"AI-powered\",\n\n // Workforce\n \"headcount reduction\", \"workforce reduction\", \"reduction in force\",\n \"workforce optimization\", \"do more with less\",\n\n // Operations / Automation\n \"back office\", \"back-office\",\n \"process automation\", \"robotic process automation\", \"RPA\",\n \"digital transformation\", \"technology transformation\",\n \"process optimization\", \"operational efficiency\",\n \"productivity improvement\",\n \"manual processes\", \"manual work\",\n \"automate\", \"automation\", \"streamline\",\n \"modernize\", \"modernization\",\n \"shared services\",\n\n // Cost\n \"cost savings\", \"cost reduction\",\n \"restructuring\",\n \"scale without\",\n];\n",[153],"64beae46-3f47-11f1-a505-42b51c65c3df","html.ts","64beaf0e-3f47-11f1-a505-42b51c65c3df",7,["D",1776971129053],"export function stripHtml(html: string): string {\n // Remove scripts and styles\n let text = html.replace(/]*>.*?<\\/script>/gis, \" \");\n text = text.replace(/]*>.*?<\\/style>/gis, \" \");\n // Strip XBRL inline tags (SEC filings have these)\n text = text.replace(/<\\/?ix:[^>]+>/gi, \" \");\n // Remove all tags\n text = text.replace(/<[^>]+>/g, \" \");\n // Unescape common HTML entities\n text = text.replace(/&/g, \"&\")\n .replace(/</g, \"<\")\n .replace(/>/g, \">\")\n .replace(/"/g, '\"')\n .replace(/'/g, \"'\")\n .replace(/ /g, \" \")\n .replace(/&#\\d+;/g, \" \")\n .replace(/&[a-zA-Z]+;/g, \" \");\n // Collapse whitespace\n text = text.replace(/\\s+/g, \" \").trim();\n return text;\n}\n",[160],"708b6160-3f47-11f1-829c-42b51c65c3df","analysis.ts","708b6200-3f47-11f1-829c-42b51c65c3df",["D",1776971148838],"// High-value neighbor words that indicate strategic intent (not boilerplate)\nconst HIGH_VALUE_NEIGHBORS = [\n \"initiative\", \"invest\", \"deploy\", \"implement\", \"launch\",\n \"reduce\", \"reduction\", \"eliminate\", \"transform\", \"strategy\",\n \"pilot\", \"roll out\", \"acquired\", \"acquisition\", \"partnership\",\n \"committed\", \"priority\", \"focus\", \"budget\", \"spend\",\n \"savings\", \"efficiency\", \"headcount\", \"workforce\",\n \"restructuring\", \"consolidat\", \"migrat\",\n];\n\n// Sections that indicate strategic commentary (higher signal)\nconst HIGH_VALUE_SECTIONS = [\n \"Item 7\", // MD&A\n \"Item 1A\", // Risk Factors\n \"Item 2\", // Properties (sometimes has ops strategy)\n];\n\nexport function scoreHit(context: string, section?: string): number {\n const lower = context.toLowerCase();\n let score = HIGH_VALUE_NEIGHBORS.filter(w => lower.includes(w)).length;\n\n // Bonus for high-signal sections\n if (section) {\n for (const s of HIGH_VALUE_SECTIONS) {\n if (section.startsWith(s)) {\n score += 2;\n break;\n }\n }\n }\n\n return score;\n}\n\nexport function normalize(s: string): string {\n return s.toLowerCase().replace(/[^a-z0-9\\s]/g, \"\").replace(/\\s+/g, \" \").trim();\n}\n\n// Find the nearest preceding \"Item X\" section header\nexport function findSection(text: string, matchIdx: number): string {\n const before = text.slice(Math.max(0, matchIdx - 10000), matchIdx);\n const sectionMatches = [...before.matchAll(/Item\\s+(\\d+[A-Z]?)\\.?\\s+([A-Z][A-Za-z\\s&']{3,60})/g)];\n if (sectionMatches.length === 0) return \"Unknown\";\n const last = sectionMatches[sectionMatches.length - 1];\n return `Item ${last[1]} - ${last[2].trim()}`;\n}\n\nexport function countByKeyword(hits: { keyword: string }[]): Record {\n const counts: Record = {};\n for (const h of hits) {\n counts[h.keyword] = (counts[h.keyword] || 0) + 1;\n }\n return counts;\n}\n",[167],"pulls","branches",1,"forks","envVars",{"_180":61,"_182":61},"connections",{"_180":61,"_181":61},"count","unfilledCount","emptyCount",{"_43":166,"_32":167,"_30":168,"_134":-5,"_41":42,"_40":150,"_136":120,"_38":205,"_13":61,"_33":138,"_35":170,"_24":25,"_140":-5,"_190":206,"_27":207},{"_43":159,"_32":160,"_30":161,"_134":-5,"_41":42,"_40":162,"_136":120,"_38":202,"_13":61,"_33":138,"_35":164,"_24":25,"_140":-5,"_190":203,"_27":204},{"_43":152,"_32":153,"_30":154,"_134":-5,"_41":42,"_40":155,"_136":120,"_38":199,"_13":61,"_33":138,"_35":157,"_24":25,"_140":-5,"_190":200,"_27":201},{"_43":148,"_32":28,"_30":31,"_134":-5,"_41":42,"_40":14,"_136":120,"_38":196,"_13":150,"_33":34,"_35":37,"_24":25,"_140":-5,"_190":197,"_27":198},{"_43":142,"_32":143,"_30":144,"_134":-5,"_41":42,"_40":145,"_136":120,"_38":193,"_13":61,"_33":138,"_35":103,"_24":25,"_140":-5,"_190":194,"_27":195},{"_43":131,"_32":132,"_30":133,"_134":-5,"_41":42,"_40":135,"_136":120,"_38":189,"_13":61,"_33":138,"_35":139,"_24":25,"_140":-5,"_190":191,"_27":192},["D",1776971136950],"children",[],[132],["D",1776974890453],[],[143],["D",1776979764369],[],[28],["D",1776971122935],[],[153],["D",1776971129053],[],[160],["D",1776971148838],[],[167],["D",1776966215049],{"_94":93,"_101":97},{"_94":93,"_95":-5,"_96":97,"_18":55,"_98":99,"_100":25},{"_13":14,"_32":84,"_43":42,"_85":44,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":45,"_92":93},"ctx",{"_77":-5,"_122":245,"_246":247,"_248":249,"_250":25,"_251":25,"_252":25,"_253":120,"_254":-5,"_255":-7,"_256":-7,"_257":61,"_258":25,"_259":25,"_260":-5,"_261":25,"_262":25,"_263":25,"_264":-5},"ENV",{"_216":217,"_218":219,"_220":84,"_221":222,"_223":224,"_225":226,"_227":228,"_229":230,"_231":232,"_233":234,"_235":236,"_237":238,"_239":240,"_241":242,"_243":244},"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"]

Diff · v15