See on Twitter

[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_184":185,"_186":187},"routes/_app.x.$handle.$name",{"_97":96,"_32":45,"_105":87,"_27":106,"_13":14,"_107":-5,"_11":108,"_109":110,"_111":112,"_113":114,"_115":116,"_20":-5,"_117":-5,"_118":-5,"_119":25,"_120":121,"_122":25,"_123":-5,"_124":-5,"_125":126},"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":-5,"_53":-5,"_18":54,"_55":-5,"_56":25,"_57":-5,"_58":-5,"_59":60,"_61":62,"_63":64,"_65":66,"_67":68,"_69":-5,"_70":71,"_72":71,"_73":74,"_75":74,"_76":74,"_77":71,"_78":71,"_79":80,"_81":82,"_83":84,"_85":25},"version",21,"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","7626d02a-0d86-11f1-97be-42dde27851f2","name","fileType","script","content","/**\n * Stripe Japan Twitter Monitor - Val.town (Built-in AI)\n *\n * Uses val.town's built-in OpenAI - NO API KEY NEEDED!\n *\n * RSS Feed: https://rss.app/feeds/9v16lB0i0vTamN2x.xml (with spam filters)\n *\n * Deploy Instructions:\n * 1. Copy this entire file\n * 2. Go to val.town → New Val → HTTP\n * 3. Paste code\n * 4. Save as \"stripeJapanMonitor\"\n * 5. Done! No secrets needed!\n *\n * Your feed: https://daniel_stripe-stripejapanmonitor.web.val.run/\n */\n\n// Val.town built-in AI - no API key needed!\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\n\nconst RSS_FEED_URL = \"https://rss.app/feeds/9v16lB0i0vTamN2x.xml\";\n\nconst CLASSIFICATION_PROMPT =\n `You are a content classifier for Stripe's Japanese market monitoring.\nDetermine if a Japanese tweet is relevant to Stripe (the payment company).\n\nRELEVANT: Stripe payment processing, APIs, products, webhooks, integrations, user feedback, company news\nIRRELEVANT: Fashion brands (STRIPE clothing), golf equipment, username greetings, spam, unrelated products\n\nReturn ONLY valid JSON: {\"relevant\": true, \"reason\": \"Brief explanation\", \"confidence\": \"high\"}`;\n\n// Val.town SQLite\nconst { sqlite } = await import(\"https://esm.town/v/std/sqlite\");\n\nexport default async function (req: Request): Promise {\n try {\n console.log(\"[Start] Initializing database...\");\n await initDB();\n console.log(\"[Start] Database initialized\");\n\n const url = new URL(req.url);\n\n // Stats endpoint\n if (url.pathname === \"/stats\") {\n return handleStats();\n }\n\n // Debug endpoint - show all tweets\n if (url.pathname === \"/debug\") {\n return handleDebug();\n }\n\n // Clear cache and re-classify\n if (url.pathname === \"/clear\") {\n await sqlite.execute({ sql: \"DELETE FROM tweets\", args: [] });\n await sqlite.execute({ sql: \"DELETE FROM metadata\", args: [] });\n console.log(\"[Clear] Database cleared\");\n return new Response(\n \"Cache cleared. Visit /?refresh=1 to re-classify tweets.\",\n {\n headers: { \"Content-Type\": \"text/plain\" },\n },\n );\n }\n\n // Force refresh with ?refresh=1\n const forceRefresh = url.searchParams.get(\"refresh\") === \"1\";\n console.log(\"[Start] Force refresh:\", forceRefresh);\n\n if (forceRefresh || await shouldRefresh()) {\n console.log(\"[Start] Starting fetch and classify...\");\n await fetchAndClassify();\n console.log(\"[Start] Fetch and classify complete\");\n }\n\n // Return filtered RSS\n const tweets = await getRelevantTweets();\n console.log(\"[Start] Returning\", tweets.length, \"relevant tweets\");\n const rss = generateRSS(tweets);\n\n return new Response(rss, {\n headers: {\n \"Content-Type\": \"application/rss+xml; charset=utf-8\",\n \"Cache-Control\": \"public, max-age=900\",\n },\n });\n } catch (error) {\n console.error(\"[Error] Fatal error:\", error);\n return new Response(`Error: ${error.message}\\n\\n${error.stack}`, {\n status: 500,\n headers: { \"Content-Type\": \"text/plain\" },\n });\n }\n}\n\nasync function initDB() {\n try {\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS tweets (\n id TEXT PRIMARY KEY,\n text TEXT NOT NULL,\n author TEXT NOT NULL,\n url TEXT NOT NULL,\n pub_date TEXT NOT NULL,\n relevant INTEGER NOT NULL,\n reason TEXT,\n confidence TEXT,\n classified_at INTEGER NOT NULL\n )\n `);\n console.log(\"[DB] tweets table created/verified\");\n\n await sqlite.execute(`\n CREATE INDEX IF NOT EXISTS idx_relevant ON tweets(relevant, classified_at DESC)\n `);\n console.log(\"[DB] index created/verified\");\n\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS metadata (\n key TEXT PRIMARY KEY,\n value TEXT NOT NULL\n )\n `);\n console.log(\"[DB] metadata table created/verified\");\n } catch (error) {\n console.error(\"[DB] Error creating tables:\", error);\n throw error;\n }\n}\n\nasync function shouldRefresh(): Promise {\n const result = await sqlite.execute({\n sql: \"SELECT value FROM metadata WHERE key = 'last_refresh'\",\n args: [],\n });\n\n if (result.rows.length === 0) return true;\n\n const lastRefresh = parseInt(result.rows[0][0] as string);\n return (Date.now() - lastRefresh) > 15 * 60 * 1000; // 15 min\n}\n\nasync function fetchAndClassify() {\n try {\n console.log(\"[Refresh] Fetching RSS feed from:\", RSS_FEED_URL);\n\n const response = await fetch(RSS_FEED_URL);\n if (!response.ok) {\n throw new Error(\n `RSS fetch failed: ${response.status} ${response.statusText}`,\n );\n }\n\n const xml = await response.text();\n console.log(\"[Refresh] RSS feed downloaded, size:\", xml.length, \"bytes\");\n\n const tweets = parseRSS(xml);\n console.log(`[Refresh] Parsed ${tweets.length} tweets from feed`);\n\n if (tweets.length === 0) {\n console.log(\"[Refresh] No tweets found in feed\");\n return;\n }\n\n let newCount = 0;\n let classifiedCount = 0;\n let relevantCount = 0;\n\n for (const tweet of tweets) {\n // Skip if already classified\n const existing = await sqlite.execute({\n sql: \"SELECT id FROM tweets WHERE id = ?\",\n args: [tweet.id],\n });\n\n if (existing.rows.length > 0) {\n console.log(`[Skip] Already classified: ${tweet.id}`);\n continue;\n }\n\n // Classify with LLM\n console.log(\n `[Classify ${\n classifiedCount + 1\n }/${tweets.length}] ${tweet.id} - @${tweet.author}`,\n );\n const classification = await classifyTweet(tweet);\n classifiedCount++;\n\n // Store in DB\n await sqlite.execute({\n sql:\n `INSERT INTO tweets (id, text, author, url, pub_date, relevant, reason, confidence, classified_at)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,\n args: [\n tweet.id,\n tweet.text,\n tweet.author,\n tweet.url,\n tweet.pubDate,\n classification.relevant ? 1 : 0,\n classification.reason,\n classification.confidence,\n Date.now(),\n ],\n });\n\n if (classification.relevant) {\n relevantCount++;\n console.log(` ✓ RELEVANT: ${classification.reason}`);\n } else {\n console.log(` ✗ filtered: ${classification.reason}`);\n }\n\n newCount++;\n\n // Rate limit: 1 request per second (val.town allows 10/min)\n await new Promise((resolve) => setTimeout(resolve, 1000));\n }\n\n console.log(\n `[Refresh] Complete: ${newCount} new tweets classified, ${relevantCount} relevant`,\n );\n\n // Update last refresh\n await sqlite.execute({\n sql:\n \"INSERT OR REPLACE INTO metadata (key, value) VALUES ('last_refresh', ?)\",\n args: [Date.now().toString()],\n });\n } catch (error) {\n console.error(\"[Refresh] Error during fetch and classify:\", error);\n throw error;\n }\n}\n\ninterface Tweet {\n id: string;\n text: string;\n author: string;\n url: string;\n pubDate: string;\n}\n\nfunction parseRSS(xml: string): Tweet[] {\n const tweets: Tweet[] = [];\n const itemRegex = /([\\s\\S]*?)<\\/item>/g;\n\n for (const match of xml.matchAll(itemRegex)) {\n const item = match[1];\n\n const title = extractTag(item, \"title\");\n const link = extractTag(item, \"link\");\n const pubDate = extractTag(item, \"pubDate\");\n const description = extractTag(item, \"description\");\n const creator = extractTag(item, \"dc:creator\");\n\n const tweetId = link.match(/status\\/(\\d+)/)?.[1];\n if (!tweetId) continue;\n\n // Clean author (remove @)\n const author = creator.replace(\"@\", \"\").trim() || \"unknown\";\n\n // Use description as text (strip HTML), fallback to title\n const text = stripHTML(description || title);\n\n tweets.push({\n id: tweetId,\n text,\n author,\n url: link,\n pubDate,\n });\n }\n\n return tweets;\n}\n\nfunction extractTag(xml: string, tag: string): string {\n const match = xml.match(\n new RegExp(\n `<${tag}>|<${tag}>([\\\\s\\\\S]*?)`,\n ),\n );\n return (match?.[1] || match?.[2] || \"\").trim();\n}\n\nfunction stripHTML(html: string): string {\n return html\n .replace(/)<[^<]*)*<\\/script>/gi, \"\")\n .replace(/<[^>]+>/g, \"\")\n .replace(/</g, \"<\")\n .replace(/>/g, \">\")\n .replace(/&/g, \"&\")\n .replace(/"/g, '\"')\n .replace(/'/g, \"'\")\n .trim();\n}\n\ninterface Classification {\n relevant: boolean;\n reason: string;\n confidence: string;\n}\n\nasync function classifyTweet(tweet: Tweet): Promise {\n try {\n // Val.town built-in OpenAI - no API key needed!\n const openai = new OpenAI();\n\n console.log(\"[LLM] Calling OpenAI for tweet:\", tweet.id);\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [\n { role: \"system\", content: CLASSIFICATION_PROMPT },\n {\n role: \"user\",\n content: `Tweet:\\nText: ${tweet.text}\\nAuthor: @${tweet.author}`,\n },\n ],\n temperature: 0.3,\n max_tokens: 150,\n });\n\n console.log(\"[LLM] Got response\");\n const response = completion.choices[0].message.content || \"{}\";\n console.log(\"[LLM] Response content:\", response);\n\n const jsonMatch = response.match(/\\{[\\s\\S]*\\}/);\n const json = jsonMatch ? JSON.parse(jsonMatch[0]) : {};\n\n return {\n relevant: json.relevant || false,\n reason: json.reason || \"Unknown\",\n confidence: json.confidence || \"low\",\n };\n } catch (error) {\n console.error(\"[LLM] Classification error:\", error);\n console.error(\"[LLM] Error details:\", {\n message: error.message,\n stack: error.stack,\n name: error.name,\n });\n return {\n relevant: false,\n reason: `Classification error: ${error.message}`,\n confidence: \"low\",\n };\n }\n}\n\nasync function getRelevantTweets(): Promise {\n const result = await sqlite.execute({\n sql: `SELECT * FROM tweets\n WHERE relevant = 1\n ORDER BY classified_at DESC\n LIMIT 50`,\n args: [],\n });\n\n return result.rows.map((row) => ({\n id: row[0],\n text: row[1],\n author: row[2],\n url: row[3],\n pubDate: row[4],\n relevant: row[5],\n reason: row[6],\n confidence: row[7],\n classifiedAt: row[8],\n }));\n}\n\nfunction generateRSS(tweets: any[]): string {\n const items = tweets.map((tweet) => `\n \n See on Twitter\n ${escapeXML(tweet.url)}\n \n ${tweet.pubDate}\n ${tweet.url}\n \n \n `).join(\"\");\n\n return `\n\n \n Stripe Japan Monitor\n https://stripe.com/jp\n Japanese tweets about Stripe payments\n ja\n ${new Date().toUTCString()}\n ${items}\n \n`;\n}\n\nfunction escapeXML(str: string): string {\n return str\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\")\n .replace(/'/g, \"'\");\n}\n\nasync function handleStats(): Promise {\n const total = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets\",\n args: [],\n });\n\n const relevant = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets WHERE relevant = 1\",\n args: [],\n });\n\n const recent = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets WHERE classified_at > ?\",\n args: [(Date.now() - 24 * 60 * 60 * 1000).toString()],\n });\n\n const lastRefresh = await sqlite.execute({\n sql: \"SELECT value FROM metadata WHERE key = 'last_refresh'\",\n args: [],\n });\n\n const stats = {\n total: total.rows[0][0],\n relevant: relevant.rows[0][0],\n relevance_rate:\n ((relevant.rows[0][0] as number) / (total.rows[0][0] as number) * 100)\n .toFixed(1) + \"%\",\n last_24h: recent.rows[0][0],\n last_refresh: lastRefresh.rows.length > 0\n ? new Date(parseInt(lastRefresh.rows[0][0] as string)).toISOString()\n : \"never\",\n };\n\n return new Response(JSON.stringify(stats, null, 2), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n}\n\nasync function handleDebug(): Promise {\n const result = await sqlite.execute({\n sql: `SELECT id, author, text, relevant, reason, confidence\n FROM tweets\n ORDER BY classified_at DESC\n LIMIT 20`,\n args: [],\n });\n\n const tweets = result.rows.map((row) => ({\n id: row[0],\n author: row[1],\n text: (row[2] as string).substring(0, 100),\n relevant: row[3] === 1,\n reason: row[4],\n confidence: row[5],\n }));\n\n return new Response(JSON.stringify(tweets, null, 2), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n}","/**\n * Stripe Japan Twitter Monitor - Val.town (Built-in AI)\n *\n * Uses val.town's built-in OpenAI - NO API KEY NEEDED!\n *\n * RSS Feed: https://rss.app/feeds/9v16lB0i0vTamN2x.xml (with spam filters)\n *\n * Deploy Instructions:\n * 1. Copy this entire file\n * 2. Go to val.town → New Val → HTTP\n * 3. Paste code\n * 4. Save as \"stripeJapanMonitor\"\n * 5. Done! No secrets needed!\n *\n * Your feed: https://daniel_stripe-stripejapanmonitor.web.val.run/\n */\n\n// Val.town built-in AI - no API key needed!\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\n\nconst RSS_FEED_URL = \"https://rss.app/feeds/9v16lB0i0vTamN2x.xml\";\n\nconst CLASSIFICATION_PROMPT =\n `You are a content classifier for Stripe's Japanese market monitoring.\nDetermine if a Japanese tweet is relevant to Stripe (the payment company).\n\nRELEVANT: Stripe payment processing, APIs, products, webhooks, integrations, user feedback, company news\nIRRELEVANT: Fashion brands (STRIPE clothing), golf equipment, username greetings, spam, unrelated products\n\nReturn ONLY valid JSON: {\"relevant\": true, \"reason\": \"Brief explanation\", \"confidence\": \"high\"}`;\n\n// Val.town SQLite\nconst { sqlite } = await import(\"https://esm.town/v/std/sqlite\");\n\nexport default async function (req: Request): Promise {\n try {\n console.log(\"[Start] Initializing database...\");\n await initDB();\n console.log(\"[Start] Database initialized\");\n\n const url = new URL(req.url);\n\n // Stats endpoint\n if (url.pathname === \"/stats\") {\n return handleStats();\n }\n\n // Debug endpoint - show all tweets\n if (url.pathname === \"/debug\") {\n return handleDebug();\n }\n\n // Clear cache and re-classify\n if (url.pathname === \"/clear\") {\n await sqlite.execute({ sql: \"DELETE FROM tweets\", args: [] });\n await sqlite.execute({ sql: \"DELETE FROM metadata\", args: [] });\n console.log(\"[Clear] Database cleared\");\n return new Response(\n \"Cache cleared. Visit /?refresh=1 to re-classify tweets.\",\n {\n headers: { \"Content-Type\": \"text/plain\" },\n },\n );\n }\n\n // Force refresh with ?refresh=1\n const forceRefresh = url.searchParams.get(\"refresh\") === \"1\";\n console.log(\"[Start] Force refresh:\", forceRefresh);\n\n if (forceRefresh || await shouldRefresh()) {\n console.log(\"[Start] Starting fetch and classify...\");\n await fetchAndClassify();\n console.log(\"[Start] Fetch and classify complete\");\n }\n\n // Return filtered RSS\n const tweets = await getRelevantTweets();\n console.log(\"[Start] Returning\", tweets.length, \"relevant tweets\");\n const rss = generateRSS(tweets);\n\n return new Response(rss, {\n headers: {\n \"Content-Type\": \"application/rss+xml; charset=utf-8\",\n \"Cache-Control\": \"public, max-age=900\",\n },\n });\n } catch (error) {\n console.error(\"[Error] Fatal error:\", error);\n return new Response(`Error: ${error.message}\\n\\n${error.stack}`, {\n status: 500,\n headers: { \"Content-Type\": \"text/plain\" },\n });\n }\n}\n\nasync function initDB() {\n try {\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS tweets (\n id TEXT PRIMARY KEY,\n text TEXT NOT NULL,\n author TEXT NOT NULL,\n url TEXT NOT NULL,\n pub_date TEXT NOT NULL,\n relevant INTEGER NOT NULL,\n reason TEXT,\n confidence TEXT,\n classified_at INTEGER NOT NULL\n )\n `);\n console.log(\"[DB] tweets table created/verified\");\n\n await sqlite.execute(`\n CREATE INDEX IF NOT EXISTS idx_relevant ON tweets(relevant, classified_at DESC)\n `);\n console.log(\"[DB] index created/verified\");\n\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS metadata (\n key TEXT PRIMARY KEY,\n value TEXT NOT NULL\n )\n `);\n console.log(\"[DB] metadata table created/verified\");\n } catch (error) {\n console.error(\"[DB] Error creating tables:\", error);\n throw error;\n }\n}\n\nasync function shouldRefresh(): Promise {\n const result = await sqlite.execute({\n sql: \"SELECT value FROM metadata WHERE key = 'last_refresh'\",\n args: [],\n });\n\n if (result.rows.length === 0) return true;\n\n const lastRefresh = parseInt(result.rows[0][0] as string);\n return (Date.now() - lastRefresh) > 15 * 60 * 1000; // 15 min\n}\n\nasync function fetchAndClassify() {\n try {\n // Step 1: Retry failed classifications\n console.log(\"[Refresh] Checking for failed classifications to retry...\");\n const failedTweets = await sqlite.execute({\n sql: `SELECT id, text, author, url, pub_date FROM tweets\n WHERE reason LIKE 'Classification error:%'\n ORDER BY pub_date ASC\n LIMIT 10`,\n args: [],\n });\n\n if (failedTweets.rows.length > 0) {\n console.log(\n `[Retry] Found ${failedTweets.rows.length} failed tweets, retrying...`,\n );\n\n for (const row of failedTweets.rows) {\n const tweet: Tweet = {\n id: row[0] as string,\n text: row[1] as string,\n author: row[2] as string,\n url: row[3] as string,\n pubDate: row[4] as string,\n };\n\n console.log(`[Retry] ${tweet.id} - @${tweet.author}`);\n const classification = await classifyTweet(tweet);\n\n // Update the existing row\n await sqlite.execute({\n sql: `UPDATE tweets\n SET relevant = ?, reason = ?, confidence = ?, classified_at = ?\n WHERE id = ?`,\n args: [\n classification.relevant ? 1 : 0,\n classification.reason,\n classification.confidence,\n Date.now(),\n tweet.id,\n ],\n });\n\n if (classification.relevant) {\n console.log(` ✓ RELEVANT: ${classification.reason}`);\n } else if (classification.reason.startsWith(\"Classification error:\")) {\n console.log(` ⚠ Still failing: ${classification.reason}`);\n } else {\n console.log(` ✗ filtered: ${classification.reason}`);\n }\n\n // Rate limit: 1 request per second (val.town allows 10/min)\n await new Promise((resolve) => setTimeout(resolve, 1000));\n }\n }\n\n // Step 2: Process new tweets from RSS feed\n console.log(\"[Refresh] Fetching RSS feed from:\", RSS_FEED_URL);\n\n const response = await fetch(RSS_FEED_URL);\n if (!response.ok) {\n throw new Error(\n `RSS fetch failed: ${response.status} ${response.statusText}`,\n );\n }\n\n const xml = await response.text();\n console.log(\"[Refresh] RSS feed downloaded, size:\", xml.length, \"bytes\");\n\n const tweets = parseRSS(xml);\n console.log(`[Refresh] Parsed ${tweets.length} tweets from feed`);\n\n if (tweets.length === 0) {\n console.log(\"[Refresh] No tweets found in feed\");\n return;\n }\n\n let newCount = 0;\n let classifiedCount = 0;\n let relevantCount = 0;\n\n for (const tweet of tweets) {\n // Skip if already classified\n const existing = await sqlite.execute({\n sql: \"SELECT id FROM tweets WHERE id = ?\",\n args: [tweet.id],\n });\n\n if (existing.rows.length > 0) {\n console.log(`[Skip] Already classified: ${tweet.id}`);\n continue;\n }\n\n // Classify with LLM\n console.log(\n `[Classify ${\n classifiedCount + 1\n }/${tweets.length}] ${tweet.id} - @${tweet.author}`,\n );\n const classification = await classifyTweet(tweet);\n classifiedCount++;\n\n // Store in DB\n await sqlite.execute({\n sql:\n `INSERT INTO tweets (id, text, author, url, pub_date, relevant, reason, confidence, classified_at)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)`,\n args: [\n tweet.id,\n tweet.text,\n tweet.author,\n tweet.url,\n tweet.pubDate,\n classification.relevant ? 1 : 0,\n classification.reason,\n classification.confidence,\n Date.now(),\n ],\n });\n\n if (classification.relevant) {\n relevantCount++;\n console.log(` ✓ RELEVANT: ${classification.reason}`);\n } else {\n console.log(` ✗ filtered: ${classification.reason}`);\n }\n\n newCount++;\n\n // Rate limit: 1 request per second (val.town allows 10/min)\n await new Promise((resolve) => setTimeout(resolve, 1000));\n }\n\n console.log(\n `[Refresh] Complete: ${newCount} new tweets classified, ${relevantCount} relevant`,\n );\n\n // Update last refresh\n await sqlite.execute({\n sql:\n \"INSERT OR REPLACE INTO metadata (key, value) VALUES ('last_refresh', ?)\",\n args: [Date.now().toString()],\n });\n } catch (error) {\n console.error(\"[Refresh] Error during fetch and classify:\", error);\n throw error;\n }\n}\n\ninterface Tweet {\n id: string;\n text: string;\n author: string;\n url: string;\n pubDate: string;\n}\n\nfunction parseRSS(xml: string): Tweet[] {\n const tweets: Tweet[] = [];\n const itemRegex = /([\\s\\S]*?)<\\/item>/g;\n\n for (const match of xml.matchAll(itemRegex)) {\n const item = match[1];\n\n const title = extractTag(item, \"title\");\n const link = extractTag(item, \"link\");\n const pubDate = extractTag(item, \"pubDate\");\n const description = extractTag(item, \"description\");\n const creator = extractTag(item, \"dc:creator\");\n\n const tweetId = link.match(/status\\/(\\d+)/)?.[1];\n if (!tweetId) continue;\n\n // Clean author (remove @)\n const author = creator.replace(\"@\", \"\").trim() || \"unknown\";\n\n // Use description as text (strip HTML), fallback to title\n const text = stripHTML(description || title);\n\n tweets.push({\n id: tweetId,\n text,\n author,\n url: link,\n pubDate,\n });\n }\n\n return tweets;\n}\n\nfunction extractTag(xml: string, tag: string): string {\n const match = xml.match(\n new RegExp(\n `<${tag}>|<${tag}>([\\\\s\\\\S]*?)`,\n ),\n );\n return (match?.[1] || match?.[2] || \"\").trim();\n}\n\nfunction stripHTML(html: string): string {\n return html\n .replace(/)<[^<]*)*<\\/script>/gi, \"\")\n .replace(/<[^>]+>/g, \"\")\n .replace(/</g, \"<\")\n .replace(/>/g, \">\")\n .replace(/&/g, \"&\")\n .replace(/"/g, '\"')\n .replace(/'/g, \"'\")\n .trim();\n}\n\ninterface Classification {\n relevant: boolean;\n reason: string;\n confidence: string;\n}\n\nasync function classifyTweet(tweet: Tweet): Promise {\n try {\n // Val.town built-in OpenAI - no API key needed!\n const openai = new OpenAI();\n\n console.log(\"[LLM] Calling OpenAI for tweet:\", tweet.id);\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [\n { role: \"system\", content: CLASSIFICATION_PROMPT },\n {\n role: \"user\",\n content: `Tweet:\\nText: ${tweet.text}\\nAuthor: @${tweet.author}`,\n },\n ],\n temperature: 0.3,\n max_tokens: 150,\n });\n\n console.log(\"[LLM] Got response\");\n const response = completion.choices[0].message.content || \"{}\";\n console.log(\"[LLM] Response content:\", response);\n\n const jsonMatch = response.match(/\\{[\\s\\S]*\\}/);\n const json = jsonMatch ? JSON.parse(jsonMatch[0]) : {};\n\n return {\n relevant: json.relevant || false,\n reason: json.reason || \"Unknown\",\n confidence: json.confidence || \"low\",\n };\n } catch (error) {\n console.error(\"[LLM] Classification error:\", error);\n console.error(\"[LLM] Error details:\", {\n message: error.message,\n stack: error.stack,\n name: error.name,\n });\n return {\n relevant: false,\n reason: `Classification error: ${error.message}`,\n confidence: \"low\",\n };\n }\n}\n\nasync function getRelevantTweets(): Promise {\n const result = await sqlite.execute({\n sql: `SELECT * FROM tweets\n WHERE relevant = 1\n ORDER BY classified_at DESC\n LIMIT 50`,\n args: [],\n });\n\n return result.rows.map((row) => ({\n id: row[0],\n text: row[1],\n author: row[2],\n url: row[3],\n pubDate: row[4],\n relevant: row[5],\n reason: row[6],\n confidence: row[7],\n classifiedAt: row[8],\n }));\n}\n\nfunction generateRSS(tweets: any[]): string {\n const items = tweets.map((tweet) => `\n \n See on Twitter\n ${escapeXML(tweet.url)}\n \n ${tweet.pubDate}\n ${tweet.url}\n \n \n `).join(\"\");\n\n return `\n\n \n Stripe Japan Monitor\n https://stripe.com/jp\n Japanese tweets about Stripe payments\n ja\n ${new Date().toUTCString()}\n ${items}\n \n`;\n}\n\nfunction escapeXML(str: string): string {\n return str\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\")\n .replace(/'/g, \"'\");\n}\n\nasync function handleStats(): Promise {\n const total = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets\",\n args: [],\n });\n\n const relevant = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets WHERE relevant = 1\",\n args: [],\n });\n\n const recent = await sqlite.execute({\n sql: \"SELECT COUNT(*) FROM tweets WHERE classified_at > ?\",\n args: [(Date.now() - 24 * 60 * 60 * 1000).toString()],\n });\n\n const lastRefresh = await sqlite.execute({\n sql: \"SELECT value FROM metadata WHERE key = 'last_refresh'\",\n args: [],\n });\n\n const stats = {\n total: total.rows[0][0],\n relevant: relevant.rows[0][0],\n relevance_rate:\n ((relevant.rows[0][0] as number) / (total.rows[0][0] as number) * 100)\n .toFixed(1) + \"%\",\n last_24h: recent.rows[0][0],\n last_refresh: lastRefresh.rows.length > 0\n ? new Date(parseInt(lastRefresh.rows[0][0] as string)).toISOString()\n : \"never\",\n };\n\n return new Response(JSON.stringify(stats, null, 2), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n}\n\nasync function handleDebug(): Promise {\n const result = await sqlite.execute({\n sql: `SELECT id, author, text, relevant, reason, confidence\n FROM tweets\n ORDER BY classified_at DESC\n LIMIT 20`,\n args: [],\n });\n\n const tweets = result.rows.map((row) => ({\n id: row[0],\n author: row[1],\n text: (row[2] as string).substring(0, 100),\n relevant: row[3] === 1,\n reason: row[4],\n confidence: row[5],\n }));\n\n return new Response(JSON.stringify(tweets, null, 2), {\n headers: { \"Content-Type\": \"application/json\" },\n });\n}","createdAt",["D",1772063437344],"branchVersion","branchId","76193190-0d86-11f1-97be-42dde27851f2","id","7618dc0e-0d86-11f1-97be-42dde27851f2","StripeJapanTwitterMonitor",["D",1771500658327],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","imageUrl","019c7588-e9d1-75f5-95f8-28956671868c","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","019c7588-e9d2-72cc-aed6-599962de2973","tursoName","7618dc0e0d8611f197be42dde27851f2","tursoHostname","7618dc0e0d8611f197be42dde27851f2-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789528252206],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_97":96,"_104":100},"org",{"_97":96,"_98":-5,"_99":100,"_18":54,"_101":102,"_103":25},"branch",{"_13":86,"_32":87,"_43":42,"_88":44,"_89":-5,"_90":-5,"_91":-5,"_92":-5,"_93":-5,"_94":45,"_95":96},"isOwned",42,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","daniel_stripe","handle","displayName","avatar","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzM5c3g4MmlLenBaZVlMWmdoVE1ZNER2RjJWNyJ9","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_43":44,"_32":45,"_38":179,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":-5,"_53":-5,"_18":54,"_55":-5,"_56":25,"_57":-5,"_58":-5,"_59":60,"_61":62,"_63":64,"_65":66,"_67":180,"_69":-5,"_70":71,"_72":71,"_73":74,"_75":74,"_76":74,"_77":71,"_78":71,"_79":181,"_81":182,"_83":183,"_85":25},"tree",[166,167,168],"files",[166,167,168],"counts",{"_154":60,"_155":156,"_157":60,"_158":159,"_160":161},"contents",[127,128,129],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_43":146,"_32":147,"_30":148,"_131":-5,"_41":42,"_40":149,"_132":25,"_38":150,"_13":151,"_33":143,"_35":152,"_24":25,"_135":-5,"_27":153},{"_43":137,"_32":138,"_30":139,"_131":-5,"_41":42,"_40":140,"_132":25,"_38":141,"_13":142,"_33":143,"_35":144,"_24":25,"_135":-5,"_27":145},{"_43":130,"_32":28,"_30":31,"_131":-5,"_41":42,"_40":14,"_132":25,"_38":133,"_13":134,"_33":34,"_35":37,"_24":25,"_135":-5,"_27":136},"c8898844-12a4-11f1-8d00-42dde27851f2","parentId","latest",["D",1772063437344],6,"label",[28],"b169bb6a-1124-11f1-8f4e-42dde27851f2","cron_weekly_digest.ts","6a4fbf4c-111d-11f1-8a10-42dde27851f2",20,["D",1771898471806],8,"interval","// Val.town Weekly Digest Cron Job\n//\n// Schedule: 0 0 * * MON (Monday midnight UTC = 9am JST)\n//\n// Generates a weekly summary of relevant Japanese Stripe tweets using GPT-4o-mini\n//\n// Deploy Instructions:\n// 1. Go to val.town → New Val → Cron\n// 2. Paste this code\n// 3. Set schedule: \"0 0 * * MON\"\n// 4. Save as \"stripeJapanWeeklyDigest\"\n\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\nimport { sqlite } from \"https://esm.town/v/std/sqlite\";\n\nconst SLACK_WEBHOOK_URL =\n \"https://hooks.slack.com/services/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\";\n\ninterface Tweet {\n id: string;\n text: string;\n author: string;\n url: string;\n pub_date: string;\n reason: string;\n confidence: string;\n}\n\nasync function getRecentTweets(): Promise {\n const sevenDaysAgo = Date.now() - (7 * 24 * 60 * 60 * 1000);\n\n const result = await sqlite.execute({\n sql: `SELECT id, text, author, url, pub_date, reason, confidence\n FROM tweets\n WHERE relevant = 1\n AND classified_at > ?\n ORDER BY pub_date DESC`,\n args: [sevenDaysAgo],\n });\n\n // SQLite returns rows as arrays, map to Tweet objects\n return result.rows.map((row) => ({\n id: row[0] as string,\n text: row[1] as string,\n author: row[2] as string,\n url: row[3] as string,\n pub_date: row[4] as string,\n reason: row[5] as string,\n confidence: row[6] as string,\n }));\n}\n\nasync function generateDigest(tweets: Tweet[]): Promise {\n if (tweets.length === 0) {\n return \"No relevant Japanese Stripe tweets this week.\";\n }\n\n const openai = new OpenAI();\n\n // Prepare tweet data for LLM with clear numbering\n const tweetSummaries = tweets.map((t, i) =>\n `[${i}] @${t.author}: ${t.text.substring(0, 200)}${\n t.text.length > 200 ? \"...\" : \"\"\n }\\n Reason: ${t.reason}`\n ).join(\"\\n\\n\");\n\n const prompt =\n `You are analyzing Japanese tweets about Stripe (the payment company) from the past week.\n\nGenerate a concise weekly digest in English using Slack's mrkdwn format.\n\nCRITICAL: You MUST cite specific tweets extensively. Every claim should link to at least one tweet source.\n\nCITATION RULES:\n- Cite tweets using double-bracketed numbers: [[0]][[1]][[2]]\n- Place citations right after the claim they support\n- EVERY claim needs at least one citation\n- Use the exact numbers from the tweet list below\n- The numbers will be converted to clickable links automatically\n\nFORMATTING RULES:\n- Use *bold* for emphasis (not **bold**)\n- Use bullet points with • or - (NOT numbered lists)\n- Do NOT use markdown headers (###, ####) - just use *bold text* instead\n- Keep it simple and readable\n\nSTRUCTURE:\n*Key Themes*\n• List 2-3 main topics/trends with citations\n\n*Notable Highlights*\n• 1-2 specific interesting observations with citations\n\n*Sentiment*\nBrief 1-sentence overall tone with citations\n\nExample:\n\"Users reported webhook issues [[2]][[5]] while others discussed Connect adoption [[1]][[8]]. A developer shared their integration experience [[0]].\"\n\nTweets from this week (cite these by number):\n${tweetSummaries}`;\n\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [\n {\n role: \"system\",\n content:\n \"You are a product analyst summarizing user feedback for Stripe's Japan team. Always cite specific sources.\",\n },\n { role: \"user\", content: prompt },\n ],\n temperature: 0.7,\n max_tokens: 1200,\n });\n\n const rawSummary = completion.choices[0].message.content ||\n \"Error generating summary\";\n\n // Convert [[0]] citations to clickable Slack mrkdwn links \n let linkedSummary = rawSummary;\n tweets.forEach((tweet, i) => {\n const citationPattern = new RegExp(`\\\\[\\\\[${i}\\\\]\\\\]`, \"g\");\n linkedSummary = linkedSummary.replace(\n citationPattern,\n `<${tweet.url}|[${i}]>`,\n );\n });\n\n return linkedSummary;\n}\n\nfunction formatSlackMessage(summary: string, tweets: Tweet[]): object {\n // Slack has 3000 char limit per text block\n const truncatedSummary = summary.length > 2800\n ? summary.substring(0, 2800) + \"...\"\n : summary;\n\n return {\n blocks: [\n {\n type: \"header\",\n text: {\n type: \"plain_text\",\n text: \"Weekly Digest: Japanese Stripe Tweets\",\n emoji: true,\n },\n },\n {\n type: \"section\",\n text: {\n type: \"mrkdwn\",\n text: `*${\n new Date().toLocaleDateString(\"en-US\", {\n month: \"long\",\n day: \"numeric\",\n year: \"numeric\",\n })\n }* | ${tweets.length} relevant tweets this week`,\n },\n },\n {\n type: \"divider\",\n },\n {\n type: \"section\",\n text: {\n type: \"mrkdwn\",\n text: truncatedSummary,\n },\n },\n {\n type: \"context\",\n elements: [\n {\n type: \"mrkdwn\",\n text:\n \"Generated by 🤖\",\n },\n ],\n },\n ],\n };\n}\n\nasync function postToSlack(message: object): Promise {\n const response = await fetch(SLACK_WEBHOOK_URL, {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify(message),\n });\n\n if (!response.ok) {\n throw new Error(\n `Slack post failed: ${response.status} ${response.statusText}`,\n );\n }\n\n console.log(\"✅ Weekly digest posted to Slack\");\n}\n\nexport default async function () {\n try {\n console.log(\"[Weekly Digest] Starting...\");\n\n const tweets = await getRecentTweets();\n console.log(\n `[Weekly Digest] Found ${tweets.length} relevant tweets from past 7 days`,\n );\n\n const summary = await generateDigest(tweets);\n console.log(\"[Weekly Digest] Generated summary\");\n\n const slackMessage = formatSlackMessage(summary, tweets);\n await postToSlack(slackMessage);\n\n console.log(\"[Weekly Digest] Complete!\");\n } catch (error) {\n console.error(\"[Weekly Digest] Error:\", error);\n throw error;\n }\n}",[138],"2493be4e-1123-11f1-833b-42dde27851f2","cron_refresh.ts","ad9d5bce-1122-11f1-b03d-42dde27851f2",17,["D",1771897806026],4,"// Val.town Cron Job - Auto-refresh every 15 minutes\n//\n// Schedule: */15 * * * * (every 15 minutes)\n//\n// Deploy Instructions:\n// 1. Go to val.town → New Val → Cron\n// 2. Paste this code\n// 3. Set schedule: */15 * * * *\n// 4. Save as \"stripeJapanMonitorRefresh\"\n//\n// This keeps the feed fresh by triggering classification of new tweets\n// from the RSS.app feed every 15 minutes.\n\nexport default async function () {\n const valUrl =\n \"https://daniel_stripe-stripejapanmonitor.web.val.run?refresh=1\";\n\n try {\n const response = await fetch(valUrl);\n console.log(\n `[Cron] Refresh triggered: ${response.status} ${response.statusText}`,\n );\n } catch (error) {\n console.error(`[Cron] Refresh failed:`, error);\n }\n}",[147],"pulls","branches",1,"forks","envVars",{"_162":164,"_165":60},"connections",{"_162":60,"_163":60},"count","unfilledCount",2,"emptyCount",{"_43":146,"_32":147,"_30":148,"_131":-5,"_41":42,"_40":149,"_132":25,"_38":176,"_13":151,"_33":143,"_35":152,"_24":25,"_135":-5,"_170":177,"_27":178},{"_43":137,"_32":138,"_30":139,"_131":-5,"_41":42,"_40":140,"_132":25,"_38":173,"_13":142,"_33":143,"_35":144,"_24":25,"_135":-5,"_170":174,"_27":175},{"_43":130,"_32":28,"_30":31,"_131":-5,"_41":42,"_40":14,"_132":25,"_38":169,"_13":134,"_33":34,"_35":37,"_24":25,"_135":-5,"_170":171,"_27":172},["D",1772063437344],"children",[],[28],["D",1771898471806],[],[138],["D",1771897806026],[],[147],["D",1771500658327],["D",1789528252206],{"_97":96,"_104":100},{"_97":96,"_98":-5,"_99":100,"_18":54,"_101":102,"_103":25},{"_13":86,"_32":87,"_43":42,"_88":44,"_89":-5,"_90":-5,"_91":-5,"_92":-5,"_93":-5,"_94":45,"_95":96},"ctx",{"_79":-5,"_124":217,"_218":219,"_220":221,"_222":25,"_223":25,"_224":25,"_225":226,"_227":-5,"_228":-7,"_229":-7,"_230":60,"_231":25,"_232":25,"_233":-5,"_234":25,"_235":25,"_236":25,"_237":-5},"ENV",{"_188":189,"_190":191,"_192":87,"_193":194,"_195":196,"_197":198,"_199":200,"_201":202,"_203":204,"_205":206,"_207":208,"_209":210,"_211":212,"_213":214,"_215":216},"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",true,"preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v21