🤖 Elegí con quién hablar

Un compañero distinto, una respuesta distinta.

${buttons}

\n ⚠️ Aviso: las respuestas se basan en material informativo proporcionado y pueden\n contener errores u omisiones; consultá siempre las fuentes oficiales.\n Los personajes son ficticios; cualquier parecido con la realidad es mera coincidencia.\n Esta página usa cookies para recordar tu personalidad elegida y tu conversación.\n

\n\n`;\n}\n\nfunction emojiOf(name: string): string {\n // Grab the trailing emoji cluster from names like \"Delegada 😠\" or \"Profe 👩🏫\".\n const m = name.match(/[\\p{Extended_Pictographic}](?:\\u{200D}[\\p{Extended_Pictographic}])*[\\u{FE0F}]?$/u);\n return m ? m[0] : \"🙂\";\n}\n\n// ---------- Chat (returning users) ----------\nfunction buildChatPage(\n history: { role: string; text: string; created_at: number }[],\n personality: string,\n): string {\n const personas = Object.entries(PERSONALITIES).map(([key, p]) =>\n ``,\n ).join(\"\");\n\n const bubbles = history\n .map((m) => {\n const side = m.role === \"user\" ? \"user\" : \"bot\";\n const when = new Date(Number(m.created_at)).toLocaleTimeString(\"es-AR\", {\n hour: \"2-digit\",\n minute: \"2-digit\",\n });\n return `

${escapeHtml(m.text)}${when}

`;\n })\n .join(\"\");\n\n const page = renderToStaticMarkup(\n React.createElement(\n \"html\",\n null,\n React.createElement(\"head\", null,\n React.createElement(\"meta\", { charSet: \"utf-8\" }),\n React.createElement(\"title\", null, \"FAQ Chatbot\"),\n React.createElement(\"meta\", { name: \"viewport\", content: \"width=device-width, initial-scale=1\" }),\n React.createElement(\"style\", null, CHAT_CSS),\n ),\n React.createElement(\"body\", null,\n React.createElement(\"div\", { id: \"app\" },\n React.createElement(\"header\", null,\n React.createElement(\"h1\", null, \"🤖 FAQ Chatbot\"),\n React.createElement(\"select\", { id: \"personality\", dangerouslySetInnerHTML: { __html: personas } }),\n React.createElement(\"button\", { id: \"clear\", className: \"clear\" }, \"🧹\"),\n ),\n React.createElement(\"div\", { id: \"messages\", dangerouslySetInnerHTML: { __html: bubbles } }),\n React.createElement(\"div\", { id: \"inputbar\" },\n React.createElement(\"input\", { id: \"input\", placeholder: \"Escribí tu pregunta…\", autocomplete: \"off\" }),\n React.createElement(\"button\", { id: \"send\" }, \"➤\"),\n ),\n ),\n React.createElement(\"script\", {\n dangerouslySetInnerHTML: {\n __html: \"window.__config = { history: \" + JSON.stringify(history) + \" };\",\n },\n }),\n React.createElement(\"script\", { dangerouslySetInnerHTML: { __html: CLIENT_JS } }),\n ),\n ),\n );\n\n return \"\\n\" + page;\n}\n\n// ---------- helpers ----------\nfunction getCookie(header: string, name: string): string | null {\n for (const part of header.split(\";\")) {\n const eq = part.indexOf(\"=\");\n if (eq === -1) continue;\n const k = part.slice(0, eq).trim();\n if (k === name) return decodeURIComponent(part.slice(eq + 1).trim());\n }\n return null;\n}\n\nfunction redirectWithCookie(path: string, sessionId: string): Response {\n const res = new Response(null, { status: 302, headers: { location: path } });\n res.headers.set(\n \"Set-Cookie\",\n `${SESSION_COOKIE}=${encodeURIComponent(sessionId)}; Path=/; HttpOnly; SameSite=Lax; Max-Age=2592000`,\n );\n return res;\n}\n\nfunction json(data: unknown, status = 200): Response {\n return new Response(JSON.stringify(data), {\n status,\n headers: { \"content-type\": \"application/json\" },\n });\n}\n\nfunction escapeHtml(s: string): string {\n return s\n .replace(/&/g, \"&\")\n .replace(//g, \">\")\n .replace(/\"/g, \""\");\n}\n\nconst CHAT_CSS = `\n * { box-sizing: border-box; }\n body { margin: 0; font-family: -apple-system, system-ui, sans-serif; background: #e5ddd5; }\n #app { display: flex; flex-direction: column; height: 100vh; height: 100dvh; max-width: 720px; margin: 0 auto; background: #efe7dd; border-left: 1px solid #d5cdc5; border-right: 1px solid #d5cdc5; }\n header { display: flex; align-items: center; gap: 10px; padding: 10px 14px; background: #075e54; color: #fff; position: sticky; top: 0; z-index: 10; flex-shrink: 0; }\n header h1 { font-size: 17px; flex: 1; margin: 0; }\n header select { padding: 6px 8px; border-radius: 6px; border: none; font-size: 14px; }\n header .clear { background: rgba(255,255,255,.2); color: #fff; border: none; border-radius: 6px; padding: 6px 10px; cursor: pointer; font-size: 15px; }\n #messages { flex: 1; overflow-y: auto; padding: 14px; display: flex; flex-direction: column; gap: 4px; }\n .row { display: flex; width: 100%; }\n .row.user { justify-content: flex-end; }\n .row.bot { justify-content: flex-start; }\n .bubble { max-width: 78%; padding: 8px 12px; border-radius: 10px; font-size: 15px; line-height: 1.4; position: relative; box-shadow: 0 1px 1px rgba(0,0,0,.08); white-space: pre-wrap; }\n .bubble.user { background: #dcf8c6; border-bottom-right-radius: 2px; }\n .bubble.bot { background: #fff; border-bottom-left-radius: 2px; }\n .bubble .time { float: right; margin-left: 8px; margin-top: 6px; font-size: 11px; color: #999; }\n .bubble.typing { padding: 14px 18px; }\n .bubble.typing span { display: inline-block; width: 8px; height: 8px; margin: 0 2px; border-radius: 50%; background: #999; animation: typing 1.2s infinite ease-in-out; }\n .bubble.typing span:nth-child(2) { animation-delay: .2s; }\n .bubble.typing span:nth-child(3) { animation-delay: .4s; }\n @keyframes typing { 0%, 60%, 100% { transform: translateY(0); opacity: .4; } 30% { transform: translateY(-4px); opacity: 1; } }\n #inputbar { display: flex; gap: 8px; padding: 10px 14px; background: #f0f0f0; border-top: 1px solid #ddd; }\n #inputbar input { flex: 1; padding: 12px 14px; font-size: 16px; border: none; border-radius: 24px; outline: none; background: #fff; }\n #inputbar button { width: 46px; height: 46px; border-radius: 50%; border: none; background: #075e54; color: #fff; font-size: 18px; cursor: pointer; }\n`;\n\nconst CLIENT_JS = `\n const history = window.__config.history || [];\n const messages = document.getElementById(\"messages\");\n const input = document.getElementById(\"input\");\n const send = document.getElementById(\"send\");\n const clearBtn = document.getElementById(\"clear\");\n const persona = document.getElementById(\"personality\");\n\n function scrollBottom() { messages.scrollTop = messages.scrollHeight; }\n\n function bubbleHTML(m) {\n const side = m.role === \"user\" ? \"user\" : \"bot\";\n const when = new Date(Number(m.created_at)).toLocaleTimeString(\"es-AR\", { hour: \"2-digit\", minute: \"2-digit\" });\n return '

' + esc(m.text) + '' + when + '

';\n }\n\n function render(hist) {\n messages.innerHTML = (hist || []).map(bubbleHTML).join(\"\");\n scrollBottom();\n }\n\n function showTyping() {\n messages.insertAdjacentHTML(\"beforeend\",\n '

');\n scrollBottom();\n }\n\n function esc(s) {\n return (s || \"\").replace(/&/g, \"&\").replace(//g, \">\");\n }\n\n async function sendMessage() {\n const text = input.value.trim();\n if (!text) return;\n input.value = \"\";\n render((window.__hist || []).concat([{ role: \"user\", text: text, created_at: Date.now() }]));\n showTyping();\n const res = await fetch(\"/api/chat\", { method: \"POST\", headers: { \"content-type\": \"application/json\" }, body: JSON.stringify({ text }) });\n const data = await res.json();\n window.__hist = data.history;\n render(data.history);\n }\n\n send.addEventListener(\"click\", sendMessage);\n input.addEventListener(\"keydown\", (e) => { if (e.key === \"Enter\") sendMessage(); });\n // Changing personality updates the session cookie, then clears the visible chat.\n persona.addEventListener(\"change\", () => {\n window.location.href = \"/choose?personality=\" + persona.value;\n });\n clearBtn.addEventListener(\"click\", async () => {\n await fetch(\"/api/clear\", { method: \"POST\" });\n window.__hist = [];\n render([]);\n });\n\n window.__hist = history;\n render(history);\n`;\n","deleted",false,"label","primeroprimera",[36],"33036cbc-cd62-4d5a-8631-7bbd5364cc74","lib.ts","f9788d16-ae16-11f1-ad57-1607ee4eb77e",49,["D",1789420944337],26,"script","// Shared FAQ + personalities logic for the FAQ chatbot.\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\n\nexport const SHEET_URL =\n \"https://docs.google.com/spreadsheets/d/e/2PACX-1vSerSBZETsDKevnovm-yTNi79xVoemUOdATL-rgxz99VoeNhG6q1L131fnjEsJr4gPMg-88jDSD7yM9/pub?output=csv\";\n\n// Cache the sheet so we don't refetch on every request.\nlet cached: { q: string; answers: string[] }[] | null = null;\nlet cachedAt = 0;\nconst CACHE_MS = 60_000;\n\nexport async function loadFaq(): Promise<{ q: string; answers: string[] }[]> {\n if (cached && Date.now() - cachedAt < CACHE_MS) return cached;\n const res = await fetch(SHEET_URL);\n const csv = await res.text();\n const rows = csv.trim().split(\"\\n\").map(splitCsv);\n\n const questions: { num: string; text: string }[] = [];\n const answersByQuestion: Record = {};\n let currentGroup: string | null = null;\n\n for (const [a, b] of rows) {\n const aTrim = (a ?? \"\").trim();\n const bTrim = (b ?? \"\").trim();\n\n const qMatch = aTrim.match(/^Pregunta\\s*(\\d+)/i);\n if (qMatch) {\n questions.push({ num: qMatch[1], text: bTrim });\n continue;\n }\n\n const rMatch = aTrim.match(/^Respuestas?\\s+a\\s+Pregunta\\s*(\\d+)/i);\n if (rMatch) {\n currentGroup = rMatch[1];\n answersByQuestion[currentGroup] = [bTrim];\n continue;\n }\n\n if (!aTrim && bTrim && currentGroup) {\n answersByQuestion[currentGroup].push(bTrim);\n }\n }\n\n cached = questions\n .map((q) => ({ q: q.text, answers: answersByQuestion[q.num] ?? [] }))\n .filter((f) => f.q && f.answers.length > 0);\n\n cachedAt = Date.now();\n return cached;\n}\n\nfunction splitCsv(line: string): string[] {\n const out: string[] = [];\n let cur = \"\";\n let inQuotes = false;\n for (const ch of line) {\n if (ch === '\"') inQuotes = !inQuotes;\n else if (ch === \",\" && !inQuotes) {\n out.push(cur);\n cur = \"\";\n } else cur += ch;\n }\n out.push(cur);\n return out;\n}\n\n// Fallback (no API key): pick the best matching question by keyword overlap,\n// return a random one of its answers.\nexport function keywordAnswer(\n question: string,\n faq: { q: string; answers: string[] }[],\n): string {\n const words = question.toLowerCase().split(/\\W+/).filter(Boolean);\n let best: { item: { q: string; answers: string[] }; score: number } | null =\n null;\n for (const item of faq) {\n const qWords = item.q.toLowerCase().split(/\\W+/).filter(Boolean);\n let score = 0;\n for (const w of words) if (qWords.includes(w)) score++;\n if (!best || score > best.score) best = { item, score };\n }\n if (!best || best.score === 0) {\n return \"Uy, no tengo eso en el bolsillo todavía. Reformulámela que capaz te doy una mano.\";\n }\n const answers = best.item.answers;\n return answers[Math.floor(Math.random() * answers.length)];\n}\n\n// ------- Personalities -------\n// Each personality shapes how the bot talks while still grounded in the FAQ.\n// `identity` gives the bot a backstory so it can answer questions about itself\n// (e.g. \"¿Sos delegada?\" → yes, because its identity says so).\nexport const PERSONALITIES: Record<\n string,\n { name: string; identity: string; system: string }\n> = {\n delegada: {\n name: \"Delegada 😠\",\n identity:\n \"Sos la delegada del curso. Si te preguntan si sos la delegada o si sos delegada, \" +\n \"respondé que sí, con seguridad y orgullo, y que como delegada sabés todo del curso.\",\n system:\n \"Sos una delegada escolar, mandona y hablas como abogada en español rioplatense (voseo). \" +\n \"Siempre respondé en oraciones largas, con un toque sarcastico o canchero \" +\n \"Dá la info del FAQ y agregale SIEMPRE una pequeña frase de acompañamiento. \" +\n \"Si no sabés, decilo con altura y como sintiendote superior.\",\n },\n docente: {\n name: \"Profe 👩‍🏫\",\n identity:\n \"Sos la profesora. Si te preguntan si sos la profe o docente, \" +\n \"respondé que sí, que sos la profe del curso y que por eso sabés todo de la materia.\",\n system:\n \"Sos un profesor claro, didáctico y entusiasta que habla en español. \" +\n \"Respondé SIEMPRE en 2-3 oraciones: primer respuesta sera una pista del FAQ, después una breve explicación \" +\n \"de por qué, y cerra con 1 consejo corto para estudiar (ej: repasar tal tema, hacer un resumen). \" +\n \"Usá un tono de profe piola, no solemne. Basate en la info del FAQ.\",\n },\n chistoso: {\n name: \"El gracioso del curso 😜\",\n identity:\n \"Sos un compañero del curso, el más gracioso y carismático de la clase. \" +\n \"Si te preguntan quién sos o si sos el chistoso, respondé que sí, que sos el compa chistoso del curso.\",\n system:\n \"Sos un compa sarcástico pero buena onda que habla en español rioplatense (voseo). \" +\n \"Respondé SIEMPRE en 1-2 oraciones: primero la respuesta del FAQ, después agregale un chiste corto, \" +\n \"un comentario irónico o un remate gracioso relacionado. \" +\n \"El humor SIEMPRE después de la data, que la info se entienda clara. No te pases de largo ni seas grosero.\",\n },\n breve: {\n name: \"El niño sabelotodo 🤓\",\n identity:\n \"Sos un compañero del curso sabelotodo, reservado y de pocas palabras. \" +\n \"Si te preguntan quién sos, respondé en pocas palabras quién sos.\",\n system:\n \"Sos ultralacónico y directo y mas que nada sabelotodo. Respondé en español con UNA frase cortísima o incluso una palabra \" +\n \"si alcanza. Nada de rodeos, nada de explicaciones, nada de saludos. \" +\n 'Basate SIEMPRE en la info del FAQ. Si no sabés, decilo en máximo 3 palabras (ej: \"no sé\").',\n },\n};\n\nexport const DEFAULT_PERSONALITY = \"delegada\";\n\n// ------- LLM layer -------\n// Uses OpenAI-compatible API (`/chat/completions`). It works with:\n// - DeepSeek (cheapest): set LLM_API_KEY to a DeepSeek key, base URL stays\n// https://api.deepseek.com (default), model defaults to deepseek-v4-flash.\n// - OpenAI: set OPENAI_API_KEY and LLM_BASE_URL=https://api.openai.com/v1,\n// LLM_MODEL=gpt-4o-mini.\n// Without any key, falls back to keywordAnswer().\nconst LLM_BASE_URL = Deno.env.get(\"LLM_BASE_URL\") ?? \"https://api.deepseek.com\";\n// default to DeepSeek's cheap flash model if the LLM key is a DeepSeek one\nconst LLM_MODEL = Deno.env.get(\"LLM_MODEL\") ?? \"deepseek-v4-flash\";\n// DeepSeek key is the primary; OpenAI key is secondary.\n// Accepts DEEPSEEK, DEEPSEEK_API_KEY, or LLM_API_KEY for the DeepSeek key.\nconst DEEPSEEK_API_KEY = Deno.env.get(\"DEEPSEEK\") ??\n Deno.env.get(\"DEEPSEEK_API_KEY\") ??\n Deno.env.get(\"LLM_API_KEY\");\nconst LLM_API_KEY = DEEPSEEK_API_KEY ?? Deno.env.get(\"OPENAI_API_KEY\");\n\n// A cheap fingerprint of the FAQ so the cache invalidates when the sheet changes.\nfunction faqSignature(faq: { q: string; answers: string[] }[]): string {\n let s = \"\";\n for (const f of faq) s += f.q + \"|\" + f.answers.join(\"^\") + \"\\n\";\n return s.slice(0, 2000);\n}\n\nasync function ensureCacheDb() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS answer_cache (\n key TEXT PRIMARY KEY, -- normalized question + personality\n faq_sig TEXT NOT NULL,\n answer TEXT NOT NULL,\n created_at INTEGER NOT NULL\n )`);\n}\n\n// Cost-saver: only hit the LLM the FIRST time a question is asked. Reuse the\n// stored answer for repeated questions until the FAQ changes.\nasync function getCachedAnswer(\n question: string,\n faq: { q: string; answers: string[] }[],\n personality: string,\n): Promise {\n await ensureCacheDb();\n const sig = faqSignature(faq);\n const key = personality + \"::\" + question.toLowerCase().trim();\n const r = await sqlite.execute({\n sql: \"SELECT answer, faq_sig FROM answer_cache WHERE key = ?\",\n args: [key],\n });\n const row = (r.rows ?? [])[0] as\n | { answer: string; faq_sig: string }\n | undefined;\n if (!row) return null;\n if (row.faq_sig !== sig) return null; // sheet changed → re-ask\n return row.answer;\n}\n\nasync function storeCached(\n question: string,\n faq: { q: string; answers: string[] }[],\n personality: string,\n answer: string,\n) {\n await ensureCacheDb();\n await sqlite.execute({\n sql:\n \"INSERT OR REPLACE INTO answer_cache (key, faq_sig, answer, created_at) VALUES (?, ?, ?, ?)\",\n args: [\n personality + \"::\" + question.toLowerCase().trim(),\n faqSignature(faq),\n answer,\n Date.now(),\n ],\n });\n}\n\nexport async function llmAnswer(\n question: string,\n faq: { q: string; answers: string[] }[],\n personality: string,\n): Promise<{ text: string; usedAI: boolean }> {\n const persona = PERSONALITIES[personality] ??\n PERSONALITIES[DEFAULT_PERSONALITY];\n\n // No API key configured → fall back to keyword matching (free).\n if (!LLM_API_KEY) {\n return { text: keywordAnswer(question, faq), usedAI: false };\n }\n\n // Reuse a prior answer for this question+personality when the FAQ unchanged.\n const hit = await getCachedAnswer(question, faq, personality);\n if (hit) {\n return { text: hit, usedAI: true };\n }\n\n const faqBlock = faq\n .map((f) => `P: ${f.q}\\nR: ${f.answers.join(\" | \")}`)\n .join(\"\\n\\n\");\n\n const res = await fetch(`${LLM_BASE_URL}/chat/completions`, {\n method: \"POST\",\n headers: {\n \"content-type\": \"application/json\",\n authorization: `Bearer ${LLM_API_KEY}`,\n },\n body: JSON.stringify({\n model: LLM_MODEL,\n temperature: 0.9,\n max_tokens: 1000,\n messages: [\n { role: \"system\", content: persona.identity },\n { role: \"system\", content: persona.system },\n {\n role: \"system\",\n content: \"Este es el FAQ de referencia. Respondé basandote en el. \" +\n \"SIEMPRE hablá en español rioplatense con voseo (vos, tenés, querés, mirá), como un argentino. \" +\n \"Sumá 1 o 2 emojis a cada respuesta para darle onda y ser más gracioso (ej: 😎 🙄 😂 🤓). \" +\n \"Si el usuario pregunta algo que NO esta en el FAQ pero es sobre vos, tu identidad o personalidad, \" +\n \"respondé usando tu identidad y tono. \" +\n \"Si la pregunta NO esta en el FAQ y tampoco es sobre tu identidad, inventá una respuesta \" +\n \"plausible y graciosa en tu personaje. \" +\n \"NUNCA asumas el género del usuario (ni varón ni mujer). Si lo amerita y no sabés su nombre, \" +\n 'preguntáselo de forma neutra (ej: \"¿Cómo te llamás?\") y usalo después para personalizar. ' +\n \"\\n\\n\" + faqBlock,\n },\n { role: \"user\", content: question },\n ],\n }),\n });\n\n if (!res.ok) {\n const err = await res.text();\n console.error(\"LLM error\", res.status, err);\n return { text: keywordAnswer(question, faq), usedAI: false };\n }\n\n const data = await res.json();\n const text = data?.choices?.[0]?.message?.content?.trim();\n if (!text) {\n console.error(\n \"LLM returned empty content\",\n JSON.stringify(data).slice(0, 500),\n );\n return { text: keywordAnswer(question, faq), usedAI: false };\n }\n await storeCached(question, faq, personality, text); // save for next time\n return { text, usedAI: true };\n}\n\n// ------- Conversation history (WhatsApp-style) -------\n// Persist messages per conversation id so a refresh keeps the thread.\nexport async function ensureDb() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS messages (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n convo TEXT NOT NULL,\n role TEXT NOT NULL, -- 'user' | 'bot'\n text TEXT NOT NULL,\n created_at INTEGER NOT NULL\n )`);\n await sqlite.execute(\n `CREATE INDEX IF NOT EXISTS idx_convo ON messages (convo, id)`,\n );\n}\n\nexport async function addMessage(convo: string, role: string, text: string) {\n await ensureDb();\n await sqlite.execute({\n sql:\n \"INSERT INTO messages (convo, role, text, created_at) VALUES (?, ?, ?, ?)\",\n args: [convo, role, text, Date.now()],\n });\n}\n\nexport async function getMessages(convo: string) {\n await ensureDb();\n const r = await sqlite.execute({\n sql:\n \"SELECT role, text, created_at FROM messages WHERE convo = ? ORDER BY id ASC LIMIT 200\",\n args: [convo],\n });\n return (r.rows ?? []) as { role: string; text: string; created_at: number }[];\n}\n\nexport async function clearMessages(convo: string) {\n await ensureDb();\n await sqlite.execute({\n sql: \"DELETE FROM messages WHERE convo = ?\",\n args: [convo],\n });\n} // ------- Sessions -------\n// A session binds a browser (via a cookie) to a chosen personality + convo.\n// A new visitor has no session → they pick a personality first.\nexport async function ensureSessionDb() {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS sessions (\n id TEXT PRIMARY KEY,\n personality TEXT NOT NULL,\n convo TEXT NOT NULL,\n created_at INTEGER NOT NULL\n )`);\n}\n\nexport async function getSession(\n id: string,\n): Promise<{ personality: string; convo: string } | null> {\n await ensureSessionDb();\n const r = await sqlite.execute({\n sql: \"SELECT personality, convo FROM sessions WHERE id = ?\",\n args: [id],\n });\n const row = (r.rows ?? [])[0] as\n | { personality: string; convo: string }\n | undefined;\n return row\n ? { personality: String(row.personality), convo: String(row.convo) }\n : null;\n}\n\nexport async function setSession(\n id: string,\n personality: string,\n convo: string,\n) {\n await ensureSessionDb();\n await sqlite.execute({\n sql:\n \"INSERT OR REPLACE INTO sessions (id, personality, convo, created_at) VALUES (?, ?, ?, ?)\",\n args: [id, personality, convo, Date.now()],\n });\n}\n// ------- Opening greeting -------\n// The greeting a character sends when picked. Instead of hardcoding \"exam on\n// Monday\", it's generated from the CURRENT FAQ content, so when the sheet\n// changes topic the greeting adapts automatically. Cached by FAQ signature so\n// it's only regenerated when the sheet changes (cost-friendly).\nconst STATIC_GREETINGS: Record = {\n delegada:\n \"¡Che, bienvenid@! 👋 Soy la Delegada, la que sabe todo de este curso. Contame, ¿qué querés saber?\",\n docente:\n \"¡Hola! 👩‍🏫 Soy la profe. Contame, ¿en qué te doy una mano?\",\n chistoso:\n \"¡¿Qué hacés?! 😜 Soy el gracioso del curso. Preguntame lo que quieras y después te cuento un chiste.\",\n breve: \"Soy el sabelotodo. 🤓 Preguntá.\",\n};\n\nexport async function generateGreeting(\n personality: string,\n faq: { q: string; answers: string[] }[],\n): Promise {\n const persona = PERSONALITIES[personality] ??\n PERSONALITIES[DEFAULT_PERSONALITY];\n\n // Reuse the answer cache (invalidates automatically when the FAQ changes).\n const cached = await getCachedAnswer(\"__greeting__\", faq, personality);\n if (cached) return cached;\n\n // No API key → static fallback.\n if (!LLM_API_KEY) {\n return STATIC_GREETINGS[personality] ?? STATIC_GREETINGS.delegada;\n }\n\n const faqBlock = faq\n .map((f) => `P: ${f.q}\\nR: ${f.answers.join(\" | \")}`)\n .join(\"\\n\\n\");\n\n try {\n const res = await fetch(`${LLM_BASE_URL}/chat/completions`, {\n method: \"POST\",\n headers: {\n \"content-type\": \"application/json\",\n authorization: `Bearer ${LLM_API_KEY}`,\n },\n body: JSON.stringify({\n model: LLM_MODEL,\n temperature: 0.9,\n max_tokens: 1000, // deepseek-flash reasons first; needs room before the answer\n messages: [\n { role: \"system\", content: persona.identity },\n {\n role: \"system\",\n content:\n \"Sos este personaje escolar y hablás en español rioplatense con voseo. \" +\n \"Este es el material del curso: \\n\\n\" + faqBlock +\n \"\\n\\nEscribí UN mensaje de bienvenida corto (1 o 2 oraciones) que te presente \" +\n \"y pregunta que quiere saber. No uses comillas ni texto entre paréntesis. Sumá 1 emoji.\",\n },\n ],\n }),\n });\n\n if (res.ok) {\n const data = await res.json();\n const text = data?.choices?.[0]?.message?.content?.trim();\n if (text) {\n await storeCached(\"__greeting__\", faq, personality, text); // save → only rewrite when sheet changes\n return text;\n }\n console.error(\n \"greeting empty content\",\n JSON.stringify(data).slice(0, 300),\n );\n }\n } catch (e) {\n console.error(\"greeting error\", e);\n }\n\n return STATIC_GREETINGS[personality] ?? STATIC_GREETINGS.delegada;\n}",[59],{"_34":58,"_13":59,"_37":60,"_39":-5,"_40":41,"_42":61,"_44":45,"_46":73,"_17":63,"_49":64,"_51":65,"_53":54,"_55":-5,"_70":74,"_18":75},{"_34":35,"_13":36,"_37":38,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":69,"_17":48,"_49":50,"_51":52,"_53":54,"_55":56,"_70":71,"_18":72},["D",1789230173942],"children",[],[36],["D",1789420944337],[],[59],"fa71ddb1-4d13-45e3-a4b2-9fb14ebe52e9",["D",1789154591216],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","FAQ chatbot reading from a Google Sheet CSV","imageUrl","userId","19ccde2f-c4b0-46a6-82af-561e6df89055","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","01a091ea-ed57-7183-b991-f2c54fa7bd7e","tursoName","fa71ddb14d1345e3a4b29fb14ebe52e9","tursoHostname","fa71ddb14d1345e3a4b29fb14ebe52e9-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789503703835],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_11":12,"_134":130},"org",{"_11":12,"_127":128,"_129":130,"_86":87,"_131":132,"_133":54},"branch",{"_17":61,"_13":16,"_34":41,"_119":76,"_120":-5,"_121":-5,"_122":-5,"_123":-5,"_124":-5,"_125":14,"_126":12},"isOwned","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","displayName","Diego Ravecca","avatar","https://lh3.googleusercontent.com/a/ACg8ocJdEvC5DSZOpC-To6ZUw6Oc_ZfdazqGtQfI9BhzAz_txuHSM_Sw=s96-c","tier","pro","canRestrictVals","profileImageUrl","adoptOrgId",{"_34":76,"_13":14,"_46":173,"_78":-5,"_79":54,"_80":81,"_82":81,"_83":84,"_85":-5,"_86":87,"_88":38,"_89":45,"_90":-5,"_91":-5,"_92":93,"_94":95,"_96":97,"_98":99,"_100":174,"_102":-5,"_103":104,"_105":104,"_106":107,"_108":107,"_109":107,"_110":104,"_111":104,"_112":175,"_114":176,"_116":177,"_118":54},[167,168],"files",[167,168],"counts",{"_157":93,"_158":159,"_160":93,"_31":161,"_162":163},[153,154],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_34":58,"_13":59,"_37":60,"_39":-5,"_40":41,"_42":61,"_44":45,"_46":62,"_17":63,"_49":64,"_51":65,"_53":54,"_55":-5,"_18":156},{"_34":35,"_13":36,"_37":38,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":47,"_17":48,"_49":50,"_51":52,"_53":54,"_55":56,"_18":155},[36],[59],"pulls","branches",1,"forks",{"_164":159,"_166":93},"connections",{"_164":93,"_165":93},"count","unfilledCount","emptyCount",{"_34":58,"_13":59,"_37":60,"_39":-5,"_40":41,"_42":61,"_44":45,"_46":73,"_17":63,"_49":64,"_51":65,"_53":54,"_55":-5,"_70":171,"_18":172},{"_34":35,"_13":36,"_37":38,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":69,"_17":48,"_49":50,"_51":52,"_53":54,"_55":56,"_70":169,"_18":170},[],[36],[],[59],["D",1789154591216],["D",1789503703835],{"_11":12,"_134":130},{"_11":12,"_127":128,"_129":130,"_86":87,"_131":132,"_133":54},{"_17":61,"_13":16,"_34":41,"_119":76,"_120":-5,"_121":-5,"_122":-5,"_123":-5,"_124":-5,"_125":14,"_126":12},"ctx",{"_112":-5,"_150":211,"_212":213,"_214":215,"_216":54,"_217":54,"_218":54,"_219":45,"_220":-5,"_221":-7,"_222":-7,"_223":93,"_224":54,"_225":54,"_226":-5,"_227":54,"_228":54,"_229":54,"_230":-5},"ENV",{"_182":183,"_184":185,"_186":16,"_187":188,"_189":190,"_191":192,"_193":194,"_195":196,"_197":198,"_199":200,"_201":202,"_203":204,"_205":206,"_207":208,"_209":210},"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"]

Code