'+sub(b.v)+'

';\n else if(b.t==='h2') h = '

'+sub(b.v)+'

';\n else if(b.t==='h3') h = '

'+sub(b.v)+'

';\n else if(b.t==='ul') h = '

'+sub(i)+'

';\n else if(b.t==='table') h = tableHtml(b.head,b.rows);\n else if(b.t==='htable') h = hintTableHtml(b);\n else if(b.t==='extable') h = '

'+esc(b.label)+'

'+\n tableHtml(b.head,b.rows)+'

';\n else if(b.t==='code') h = codeHtml(b.v,'');\n else if(b.t==='link') h = '

'+esc(b.label)+'

';\n else if(b.t==='download') h = dlHtml(b);\n else if(b.t==='callout') h = boxHtml('callout',b.label,[b.text]);\n else if(b.t==='warn') h = boxHtml('warn',b.label,[b.text]);\n else if(b.t==='example') h = boxHtml('example',b.label,b.lines);\n else if(b.t==='task') h = boxHtml('task', b.label+' · worth up to '+b.points+' points', b.lines);\n else if(b.t==='goal') h = boxHtml('goal','End goal', b.lines);\n else if(b.t==='hint') h = hintHtml(b);\n else if(b.t==='test') h = testHtml(b);\n else if(b.t==='scoreboard') h = scoreboardHtml();\n return h;\n}\nfunction tableHtml(head,rows){\n var h='

';\n head.forEach(function(c){ h+=''; });\n h+='';\n rows.forEach(function(r){\n h+='';\n r.forEach(function(c){ h+=''; });\n h+='';\n });\n return h+'

'+sub(c)+'
'+(c?sub(c):' ')+'

';\n}\nfunction hintTableHtml(b){\n var h='

';\n b.head.forEach(function(c){ h+=''; });\n h+='';\n b.rows.forEach(function(r){\n var locked = r.hint && state.opened[r.hint.id]===undefined;\n h+='';\n if(locked){\n h+='';\n h+='';\n } else {\n r.cells.forEach(function(c){ h+=''; });\n }\n h+='';\n });\n return h+'

'+sub(c)+'
'+sub(r.cells[0])+'		'+(c?sub(c):' ')+'

';\n}\nfunction codeHtml(v,label){\n var id='c'+Math.random().toString(36).slice(2,9);\n return (label?'

'+esc(label)+'

':'')+\n '

'+\n '

'+esc(sub(v))+'

';\n}\nfunction boxHtml(kind,label,lines){\n var h='

'+sub(label)+'

';\n (lines||[]).forEach(function(l){ h+='

'+sub(l)+'

'; });\n return h+'

';\n}\nfunction dlHtml(b){\n if(!b.href) return '

⇩

'+\n esc(b.label)+'

Link coming shortly

Soon

';\n return '

'+esc(b.cta||'Download')+' →

';\n}\nfunction kindLabel(k){\n if(k==='scaffold') return 'hint';\n if(k==='steps') return 'step-by-step';\n if(k==='solution') return 'full solution';\n return k;\n}\nfunction hintHtml(b){\n var open = state.opened[b.id] !== undefined;\n var h='

'+\n '';\n if(open){\n h+='

';\n (b.lines||[]).forEach(function(l){ h+='

'+sub(l)+'

'; });\n if(b.code) h+=codeHtml(b.code,'');\n if(b.table) h+=tableHtml(b.table.head,b.table.rows);\n if(b.codes) b.codes.forEach(function(c){ h+=codeHtml(c.v,c.label); });\n if(b.table2) h+=tableHtml(b.table2.head,b.table2.rows);\n if(b.divider2) h+='

'+sub(b.divider2)+'

';\n if(b.table3) h+=tableHtml(b.table3.head,b.table3.rows);\n if(b.codes2) b.codes2.forEach(function(c){ h+=codeHtml(c.v,c.label); });\n if(b.table4) h+=tableHtml(b.table4.head,b.table4.rows);\n if(b.guide) h+=dlHtml({href:b.guide.href,label:b.guide.label,note:'Step-by-step walkthrough'});\n h+='

';\n }\n return h+'

';\n}\nfunction testHtml(b){\n var done = state.tests.indexOf(b.id) >= 0;\n return ''+\n ''+\n ''+sub(b.label)+''+\n ''+sub(b.text)+''+\n ''+(done?'+10 earned':'+10')+'';\n}\nfunction scoreboardHtml(){\n return '

Your score

'+\n '

'+total()+'

'+\n '

Task points'+taskPoints()+' / '+S.maxTask+'

'+\n '

Verification'+testPoints()+' / '+S.maxTest+'

'+\n '

Hints opened'+Object.keys(state.opened).length+'

'+\n '

';\n}\n\nfunction renderNav(){\n var groups=[], seen={};\n S.sections.forEach(function(s,i){\n if(!seen[s.group]){ seen[s.group]=1; groups.push(s.group); }\n });\n var h='';\n groups.forEach(function(g){\n h+='

'+esc(g)+'

';\n S.sections.forEach(function(s,i){\n if(s.group!==g) return;\n var done = s.taskNo ? taskTouched(s) : false;\n h+='';\n });\n });\n document.getElementById('nav').innerHTML = h;\n}\nfunction taskTouched(s){\n var t=false;\n (s.blocks||[]).forEach(function(b){\n if(b.t==='hint' && state.opened[b.id]!==undefined) t=true;\n if(b.t==='htable') (b.rows||[]).forEach(function(r){\n if(r.hint && state.opened[r.hint.id]!==undefined) t=true;\n });\n });\n return t;\n}\n\nfunction renderSection(keepScroll){\n var y = window.scrollY;\n var s = S.sections[state.idx];\n var h = '

'+sub(s.title)+'

';\n s.blocks.forEach(function(b){ h += blockHtml(b); });\n h += '';\n document.getElementById('content').innerHTML = h;\n renderNav();\n renderScore();\n if(keepScroll) window.scrollTo(0,y); else window.scrollTo(0,0);\n try{ localStorage.setItem('ea_idx', String(state.idx)); }catch(e){}\n}\n\n/* ---------- events ---------- */\ndocument.addEventListener('click', function(e){\n var go = e.target.closest('[data-go]');\n if(go){ state.idx = parseInt(go.getAttribute('data-go'),10); renderSection(); return; }\n\n var tog = e.target.closest('[data-toggle]');\n if(tog){\n var id = tog.getAttribute('data-toggle');\n var wrap = tog.closest('[data-hint]');\n if(state.opened[id]===undefined){\n state.opened[id] = parseInt(wrap.getAttribute('data-cost'),10);\n save();\n }\n renderSection(true);\n return;\n }\n\n var fill = e.target.closest('[data-fill]');\n if(fill){\n var fid = fill.getAttribute('data-fill');\n if(state.opened[fid]===undefined){\n state.opened[fid] = parseInt(fill.getAttribute('data-cost'),10);\n save();\n }\n renderSection(true);\n return;\n }\n\n var cp = e.target.closest('[data-copy]');\n if(cp){\n var pre = document.getElementById(cp.getAttribute('data-copy'));\n navigator.clipboard.writeText(pre.textContent).then(function(){\n cp.textContent='Copied'; cp.classList.add('ok');\n setTimeout(function(){ cp.textContent='Copy'; cp.classList.remove('ok'); },1600);\n });\n return;\n }\n});\n\ndocument.addEventListener('change', function(e){\n var lab = e.target.closest('[data-test]');\n if(!lab) return;\n var id = lab.getAttribute('data-test');\n var i = state.tests.indexOf(id);\n if(e.target.checked && i<0) state.tests.push(id);\n if(!e.target.checked && i>=0) state.tests.splice(i,1);\n lab.classList.toggle('done', e.target.checked);\n lab.querySelector('.test-p').textContent = e.target.checked ? '+10 earned' : '+10';\n renderScore();\n save();\n});\n\n/* ---------- gate ---------- */\nfunction updatePreview(){\n var f=document.getElementById('gf').value, l=document.getElementById('gl').value;\n var ini=initialsFrom(f,l);\n var el=document.getElementById('prev');\n el.innerHTML = ini.length===2\n ? 'Your tools will be named '+esc(ini)+'_enterpriseair_getflights'\n : '';\n document.getElementById('goBtn').disabled = !(ini.length===2 && document.getElementById('ge').value.indexOf('@')>0);\n}\n\nfunction startLab(user){\n state.user = user;\n try{ localStorage.setItem('ea_user', JSON.stringify(user)); }catch(e){}\n document.getElementById('gate').style.display='none';\n document.getElementById('app').style.display='flex';\n document.getElementById('railUser').innerHTML =\n ''+esc(user.firstName+' '+user.lastName)+''+esc(user.email)+\n ''+esc(user.initials)+'_';\n var savedIdx = 0;\n try{ savedIdx = parseInt(localStorage.getItem('ea_idx')||'0',10)||0; }catch(e){}\n state.idx = Math.min(savedIdx, S.sections.length-1);\n renderSection();\n}\n\nfunction boot(){\n ['gf','gl','ge'].forEach(function(id){\n document.getElementById(id).addEventListener('input', updatePreview);\n });\n document.getElementById('goBtn').addEventListener('click', function(){\n var user = {\n firstName: document.getElementById('gf').value.trim(),\n lastName: document.getElementById('gl').value.trim(),\n email: document.getElementById('ge').value.trim().toLowerCase(),\n initials: initialsFrom(document.getElementById('gf').value, document.getElementById('gl').value)\n };\n fetch('/api/progress?email='+encodeURIComponent(user.email))\n .then(function(r){ return r.json(); })\n .then(function(d){\n if(d && d.found){ state.opened = d.opened||{}; state.tests = d.tests||[]; }\n startLab(user);\n save();\n })\n .catch(function(){ startLab(user); });\n });\n\n var stored = null;\n try{ stored = JSON.parse(localStorage.getItem('ea_user')||'null'); }catch(e){}\n if(stored && stored.email){\n fetch('/api/progress?email='+encodeURIComponent(stored.email))\n .then(function(r){ return r.json(); })\n .then(function(d){\n if(d && d.found){ state.opened=d.opened||{}; state.tests=d.tests||[]; }\n startLab(stored);\n })\n .catch(function(){ startLab(stored); });\n }\n}\nboot();\n`;\n\n// ---------------------------------------------------------------------------\nfunction participantPage(): string {\n const payload = JSON.stringify({\n sections: SECTIONS,\n maxTask: MAX_TASK_POINTS,\n maxTest: MAX_TEST_POINTS,\n testPoints: TEST_POINTS,\n });\n return `\n\n\n\nEnterpriseAir Voice Agent Lab\n\n\n\n\n\n\n

Enterprise Bot · AIDA

EnterpriseAir Voice Agent Lab

First name

Last name

\n \n

\n\n\n\n\n\n`;\n}\n\nfunction instructorPage(rows: any[], keySet: boolean): string {\n const scored = rows.map((r) => ({ ...r, ...scoreOf(r.opened, r.tests) }))\n .sort((a, b) => b.total - a.total);\n const avg = scored.length\n ? Math.round(scored.reduce((s, r) => s + r.total, 0) / scored.length)\n : 0;\n const body = scored.length\n ? `

\n \n ` +\n scored.map((r) =>\n `\n \n \n \n \n \n \n `\n ).join(\"\") +\n `

Participant	Prefix	Score	Task points	Verification	Progress	Last seen
${esc(r.first_name)} ${\n esc(r.last_name)\n } ${\n esc(r.email)\n }	`${esc(r.initials)}_`	${r.total}	${r.taskPoints} / ${MAX_TASK_POINTS} ${r.hintsOpened} pts of hints	${r.testsDone} / 7 +${r.testPoints} pts		${ago(r.updated_at)}

`\n : `

No one has started the lab yet. Participants appear here as soon as they enter their name.

`;\n\n return `\n\n\nLab Dashboard\n\n\n

EnterpriseAir Lab — instructor view

Live scores. Refreshes every 30 seconds.

\n ${\n keySet\n ? \"\"\n : `

This dashboard is unprotected

Set an environment variable named LAB_KEY on this val, then open this page as /instructor?key=YOUR_KEY. Until then anyone with the URL can see participant names.

`\n }\n

Participants

${scored.length}

Average score

${avg}

Finished testing

${\n scored.filter((r) => r.testsDone === 7).length\n }

Max possible

${\n MAX_TASK_POINTS + MAX_TEST_POINTS\n }

\n ${body}\n

\n\n`;\n}\n\n// ---------------------------------------------------------------------------\nexport default async function (req: Request): Promise {\n await ensureSchema();\n const url = new URL(req.url);\n const path = url.pathname;\n const LAB_KEY = Deno.env.get(\"LAB_KEY\") || \"\";\n\n if (req.method === \"OPTIONS\") {\n return new Response(null, {\n status: 204,\n headers: {\n ...JSON_HEADERS,\n \"Access-Control-Allow-Methods\": \"GET,POST,OPTIONS\",\n \"Access-Control-Allow-Headers\": \"Content-Type\",\n },\n });\n }\n\n // --- save progress\n if (path === \"/api/progress\" && req.method === \"POST\") {\n let b: any = {};\n try {\n b = await req.json();\n } catch {\n return new Response(JSON.stringify({ ok: false }), {\n status: 400,\n headers: JSON_HEADERS,\n });\n }\n const email = String(b.email || \"\").trim().toLowerCase();\n if (!email) {\n return new Response(JSON.stringify({ ok: false }), {\n status: 400,\n headers: JSON_HEADERS,\n });\n }\n const now = Date.now();\n await sqlite.execute({\n sql:\n `insert into lab_progress (email, first_name, last_name, initials, opened, tests, started_at, updated_at)\n values (:e,:f,:l,:i,:o,:t,:n,:n)\n on conflict(email) do update set\n first_name=:f, last_name=:l, initials=:i, opened=:o, tests=:t, updated_at=:n`,\n args: {\n e: email,\n f: String(b.firstName || \"\"),\n l: String(b.lastName || \"\"),\n i: String(b.initials || \"\"),\n o: JSON.stringify(b.opened || {}),\n t: JSON.stringify(b.tests || []),\n n: now,\n },\n });\n return new Response(JSON.stringify({ ok: true }), {\n headers: JSON_HEADERS,\n });\n }\n\n // --- load one participant\n if (path === \"/api/progress\" && req.method === \"GET\") {\n const email = (url.searchParams.get(\"email\") || \"\").trim().toLowerCase();\n const res = await sqlite.execute({\n sql: `select * from lab_progress where email = :e`,\n args: { e: email },\n });\n const row = (res.rows as any[])[0];\n if (!row) {\n return new Response(JSON.stringify({ found: false }), {\n headers: JSON_HEADERS,\n });\n }\n return new Response(\n JSON.stringify({\n found: true,\n opened: JSON.parse(row.opened || \"{}\"),\n tests: JSON.parse(row.tests || \"[]\"),\n }),\n { headers: JSON_HEADERS },\n );\n }\n\n // --- instructor\n if (path === \"/instructor\" || path === \"/api/all\") {\n if (LAB_KEY && url.searchParams.get(\"key\") !== LAB_KEY) {\n return new Response(\"Not found\", { status: 404 });\n }\n const res = await sqlite.execute(\n `select * from lab_progress order by updated_at desc`,\n );\n const rows = res.rows as any[];\n if (path === \"/api/all\") {\n return new Response(\n JSON.stringify(\n rows.map((r) => ({\n name: r.first_name + \" \" + r.last_name,\n email: r.email,\n initials: r.initials,\n ...scoreOf(r.opened, r.tests),\n })),\n null,\n 2,\n ),\n { headers: JSON_HEADERS },\n );\n }\n return new Response(instructorPage(rows, !!LAB_KEY), {\n headers: { \"Content-Type\": \"text/html; charset=utf-8\" },\n });\n }\n\n // --- participant lab\n if (path === \"/\" || path === \"\") {\n return new Response(participantPage(), {\n headers: { \"Content-Type\": \"text/html; charset=utf-8\" },\n });\n }\n\n return new Response(\"Not found\", { status: 404 });\n}",2,["D",1789493606974],"branchVersion","label",[18],"pulls","branches","forks","envVars",{"_180":26,"_182":26},"connections",{"_180":26,"_181":26},"count","unfilledCount","emptyCount",{"_34":190,"_13":191,"_62":192,"_44":-5,"_36":37,"_170":129,"_185":103,"_38":193,"_40":194,"_64":65,"_166":195,"_66":24,"_171":-5,"_187":196,"_17":197},{"_34":165,"_13":18,"_62":63,"_44":-5,"_36":37,"_170":84,"_185":103,"_38":186,"_40":168,"_64":78,"_166":167,"_66":24,"_171":14,"_187":188,"_17":189},"latest",["D",1789493606974],"children",[],[18],"ac4d03a2-07af-4049-b4c7-01ad336618e6","lab_content.ts","176c8720-b12b-11f1-875c-1607ee4eb77e",["D",1789538492390],8,"/**\n * EnterpriseAir Voice Agent lab - content.\n *\n * Every occurrence of \"xx_\" is replaced in the browser with the signed-in\n * participant's initials, so nobody has to remember to do it by hand.\n *\n * Block types the renderer understands:\n * p, h2, h3, ul, ol, table, code, link, dl\n * callout {label, text} warn {label, text} example {label, lines}\n * task {label, points, lines}\n * hint {id, kind, cost, label, lines?, code?} kind: scaffold|nudge|steps|solution\n * test {id, label, text}\n */\n\nexport const BASE = \"https://enterpriseair-api.val.run\";\n\nexport const LINKS = {\n aida: \"https://demo-chatbot2go.enterprisebot.co/signin\",\n docs:\n \"https://enterprisebot1.sharepoint.com/:f:/g/IgAu7oatnb52QKlAVzB5HtzyAVoyuiqW3rj9_TtUgfGOmEU?e=85ufAx\",\n workbook:\n \"https://enterprisebot1.sharepoint.com/:w:/g/IQC_CDmXUnp9RqAGulqmu_2iAZmcUGS7he21F8UV2iA8-xo?e=Ga342o\",\n workbookDone:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQDeFOLCbC0MR4dyXnHJ3TDeATTyvmDpxJU3E_wxAcSKHck?e=lO1WmG\",\n guideKb:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQBIzJflSbwGQLh-Sc_589B5AfH3dSNBwEdVlhOVR7CDclU?e=s9FcJS\",\n guideAgent:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQC5BlvCI-6JRKnk6bJPWhLtAUjfp7AKXJRdkztLsTzDbwI?e=gjrOlx\",\n guideFlights:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQCtWgFUd6mXTLudBUpBKgOjAYNhNAKPrJROra3FJzigDWk?e=PtHlQh\",\n guideChain:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQAaJaj4y9-4SIajBeBfdvcKAYra8brqycOavHI8F8V5X1k?e=RK9UJz\",\n guidePublish:\n \"https://enterprisebot1.sharepoint.com/:b:/g/IQDZaOKw1BmoTIzHF-LamtdiAdOhF60OKaKIRVGXpUH-t7M?e=w6De76\",\n valtown: \"https://www.val.town/x/enterprisebotpen/enterpriseair-api\",\n csvBookings:\n \"https://enterprisebot1.sharepoint.com/:x:/g/IQAADBbXPjJCRLJzXYqmia7pAVvHXrSHanzdw8lfbj0UKZc?e=e9crkp\",\n csvFlights:\n \"https://enterprisebot1.sharepoint.com/:x:/g/IQAciu8yfZd0SpZ9jWHvhpr_AZ1TOpchtEApv2Sw5Wpqy_0?e=fnQm3W\",\n csvStatus:\n \"https://enterprisebot1.sharepoint.com/:x:/g/IQD0fToToG3bRp2PNadfXuqAAcpBkFI1A5CLvQ7MvB2tk8A?e=569CO8\",\n};\n\nconst csv = (href: string, label: string) =>\n `${label}`;\n\nconst INS_ROLE =\n `You are EnterpriseAir's virtual voice assistant, helping customers with three main tasks: checking scheduled flights between two locations, checking the live status of an existing booking, and answering general questions about EnterpriseAir policies.\n\nDo not wait for the customer to speak first. Only respond to clear, intentional speech; ignore background noise or silence.`;\n\nconst INS_SCHEDULE = `Flight Schedule Inquiries:\n- Identify the origin and destination from the customer's request, whether given as city names or airport codes.\n- Call the xx_enterpriseair_getflights tool with these values.\n- Read back matching flights naturally, including flight number, departure and arrival times, and operating days.\n- If no flights match, inform the customer plainly.`;\n\nconst INS_STATUS = `Booking Status or Flight Delay Inquiries:\n- Inform the customer you need to verify their booking first.\n- Request their booking reference, then their last name, one at a time.\n- Use the xx_enterpriseair_verifybooking tool with these details.\n- If the verification fails, ask the user to spell their last name. Try again with the same booking reference and the last name.\n- If authenticated: true, immediately call xx_enterpriseair_getstatus with the returned token.\n- Relay the flight status naturally: on time, delayed (including delay duration and new departure time if available), or cancelled.\n- If authenticated: false, calmly inform the customer the details did not match and invite them to try again without specifying which detail was incorrect.`;\n\nconst INS_POLICY =\n `General Policy Questions (baggage, check-in, cancellations, fares, refund policy, routes, etc.):\n- Answer only using the EnterpriseAir knowledge base attached to this agent. Do not guess or invent information.\n- If the knowledge base lacks the information, say you don't have that information and offer to connect the customer to a human agent.`;\n\nconst INS_RULES = `Hard Rules:\n- Never invent flight times, statuses, or policy details; only state what tools or the knowledge base provide.\n- Never mention tools, APIs, or internal systems to the customer.\n- If a tool call fails, apologize briefly and suggest trying again or contacting support.\n- Explaining policies is always allowed if covered by the knowledge base.\n- Decline any requests to perform actions outside the three supported areas (e.g., booking new flights, issuing refunds, changing or cancelling reservations). Inform the customer this assistant cannot perform those actions and offer to connect them to a human agent.\n\n# Output Format\nRespond naturally and conversationally in clear, polite English. Use complete sentences. Provide all requested information succinctly and clearly. When reading flight or status details, include all relevant data as specified. Do not include any technical or internal details in responses.`;\n\nexport const INSTRUCTIONS_FULL = [\n INS_ROLE,\n INS_SCHEDULE,\n INS_STATUS,\n INS_POLICY,\n INS_RULES,\n].join(\"\\n\\n\");\n\nconst WELCOME_MSG =\n `Welcome to EnterpriseAir. I'm your virtual assistant. I can check flight schedules, look up a booking's status, or answer questions about our policies. How can I help you today?`;\n\nconst VOICE_RULES = `Voice Response guidelines:\n\n- You are a voice-first assistant, so always respond in a way that sounds natural, clear, and conversational when spoken aloud.\n- Keep responses concise and conversational (1-3 sentences max)\n- Use natural, spoken language patterns\n- Avoid complex punctuation or formatting\n- Speak in a warm, professional tone\n- Ask one question at a time if clarification is needed\n- Use \"um\" or \"uh\" sparingly, prefer natural pauses\n- Pronounce flight numbers and gate letters clearly, spelling out letters where helpful (say \"E A seven zero one\" not \"EA701\")\n- Pronounce times in natural spoken form (say \"six oh five in the morning\" not \"06:05\")\n- When describing a delay, state the original time, the length of the delay, and the new time, in that order\n- End responses with clear conversation cues when appropriate`;\n\nconst DESC_GETFLIGHTS =\n \"Use this tool when the customer asks which flights are available between two places, what time they depart or arrive, or which days a route operates. Pass the origin and destination as airport codes (e.g. BLR) or city names (e.g. Bengaluru). Do not use this tool to check the status of a specific booking - use xx_enterpriseair_verifybooking and xx_enterpriseair_getstatus for that.\";\nconst DESC_VERIFYBOOKING =\n \"Use this tool to verify a customer's booking before discussing flight status. Call it once you have both the booking reference and the last name on the booking. The last name does not need to be an exact match - the backend accepts small typos and near-misses, since it's captured by voice. Returns authenticated true or false, and if true, a token, the flight number and the travel date.\";\nconst DESC_GETSTATUS =\n \"Use this tool to retrieve the live status of a flight after xx_enterpriseair_verifybooking has returned authenticated true. Pass the token value from that response as the token parameter. Returns status (On Time, Delayed or Cancelled), delay minutes, a new departure time if delayed, and the gate.\";\nconst DESC_KB =\n \"Use this to answer general questions about EnterpriseAir policies, baggage, routes, and check-in procedures that are not about a specific flight's live status.\";\n\nexport const TASK_COUNT = 6;\nexport const TASK_POINTS = 100;\nexport const TEST_POINTS = 10;\n\nexport const SECTIONS = [\n {\n id: \"start\",\n nav: \"How this lab works\",\n group: \"Orientation\",\n title: \"How this lab works\",\n blocks: [\n {\n t: \"p\",\n v: \"Self-paced build lab. You'll plan a voice agent in a workbook, then build it in AIDA and test it.\",\n },\n {\n t: \"p\",\n v: \"There are six tasks. Each starts with just a description of the task. Help is available, but opening each hint will cost you points.\",\n },\n {\n t: \"table\",\n head: [\"Hint level\", \"What you get\", \"Cost\"],\n rows: [\n [\n \"Hint\",\n \"One piece of the answer: a single row of a table, a single tool, a single prompt.\",\n \"-10 pts\",\n ],\n [\"Nudge\", \"A conceptual pointer, not the answer.\", \"-10 pts\"],\n [\"Step-by-step\", \"The exact steps.\", \"-25 pts\"],\n [\n \"Full solution\",\n \"The exact values plus the PDF walkthrough for that step.\",\n \"-50 pts\",\n ],\n ],\n },\n {\n t: \"p\",\n v: \"Each task starts at 100 points. Your score updates the moment you open a hint, and it saves automatically, so you can close this page and pick it up later on any machine.\",\n },\n {\n t: \"callout\",\n label: \"Verification points:\",\n text:\n \"The testing script at the end is worth bonus points rather than penalties: 10 per scenario you can demonstrate working, up to 70. They show as available in the bar above until you tick them.\",\n },\n {\n t: \"callout\",\n label: \"Your names are already personalised:\",\n text:\n \"Everyone builds in the same demo environment, so every name you create carries your initials. This page has already substituted them for you - wherever you see a tool or knowledge base name, it is yours to create, exactly as written.\",\n },\n ],\n },\n {\n id: \"case\",\n nav: \"The case\",\n group: \"Orientation\",\n title: \"The case: EnterpriseAir\",\n blocks: [\n {\n t: \"p\",\n v: \"EnterpriseAir is a mid-size carrier based in Bengaluru. It runs 16 domestic rotations a day across India, plus long-haul services to Dubai, Singapore, London, Frankfurt, New York and Male, on a fleet of A320neo, A321neo and 787-9 aircraft.\",\n },\n { t: \"p\", v: \"Three questions dominate their contact centre:\" },\n {\n t: \"ul\",\n v: [\n \"What flies between two cities.\",\n \"Whether a specific booking is delayed.\",\n \"What the baggage or cancellation rules are.\",\n ],\n },\n {\n t: \"p\",\n v: \"EnterpriseAir has bought Enterprise Bot solution. You are on the team implementing it. Phase one is a voice agent that answers those three questions on the phone, around the clock, and hands everything else to a human.\",\n },\n { t: \"h2\", v: \"What the client handed over\" },\n {\n t: \"ul\",\n v: [\n \"Four policy documents: Conditions of Carriage, Baggage Policy, FAQ, Route Network & Destinations.\",\n \"API access to their schedule and operations systems, documented in the next section.\",\n ],\n },\n { t: \"h2\", v: \"Learning objectives\" },\n { t: \"p\", v: \"By the end of this lab, you will be able to:\" },\n {\n t: \"ul\",\n v: [\n \"Plan an agent's jobs, tools and prompts before touching the platform.\",\n \"Build a DocBrain knowledge base and attach it to an agent.\",\n \"Build a Realtime Voice agent in AIDA from scratch.\",\n \"Configure an API tool, including which parameters are required and which travel in the query string.\",\n \"Design a two-tool verify-then-fetch chain using a token handoff.\",\n \"Write an instructions prompt that routes between capabilities without contradictions.\",\n \"Test an agent against a script, including edge cases.\",\n ],\n },\n ],\n },\n {\n id: \"backend\",\n nav: \"The backend\",\n group: \"Orientation\",\n title: \"The backend\",\n blocks: [\n {\n t: \"p\",\n v: \"The backend is already built, deployed and running. You are consuming it like any third-party API. This section is your reference for the whole lab.\",\n },\n { t: \"h3\", v: \"Base URL\" },\n { t: \"code\", v: BASE },\n {\n t: \"link\",\n href: LINKS.valtown,\n label: \"View the backend source on val.town\",\n },\n { t: \"h3\", v: \"What's in the database\" },\n {\n t: \"p\",\n v: \"The tables themselves sit inside the backend and can't be browsed from the outside, so each one is exported below. Open them to see the routes that exist and the bookings you can test against, before you start calling anything.\",\n },\n {\n t: \"table\",\n head: [\"Table\", \"What it holds\", \"Data\"],\n rows: [\n [\n \"flights\",\n \"The recurring published timetable: flight number, route, times, aircraft, and the days each route operates. This is what /flights reads.\",\n csv(LINKS.csvFlights, \"flights.csv\"),\n ],\n [\n \"bookings\",\n \"Test bookings: booking reference, passenger name, flight number, and how many days from today the trip is. Take your test booking references from here.\",\n csv(LINKS.csvBookings, \"bookings.csv\"),\n ],\n [\n \"flight_status\",\n \"The status of a specific flight on a specific day, plus delay minutes and gate. This is what /flightstatus reads back.\",\n csv(LINKS.csvStatus, \"flight_status.csv\"),\n ],\n [\n \"verify_tokens\",\n \"Short-lived tokens issued by /verifybooking, each tied to the flight and travel day it was issued for. Written by the backend as you test; nothing to inspect up front.\",\n \"-\",\n ],\n ],\n },\n { t: \"h2\", v: \"GET /flights\" },\n {\n t: \"p\",\n v: \"Looks up the published schedule between two places. No verification: a timetable is public information.\",\n },\n {\n t: \"table\",\n head: [\"Parameter\", \"Required\", \"Sent as\", \"Notes\"],\n rows: [\n [\n \"origin\",\n \"Yes\",\n \"Query string\",\n \"Airport code or city name. The backend returns 400 if it's missing.\",\n ],\n [\n \"destination\",\n \"Yes\",\n \"Query string\",\n \"Airport code or city name. The backend returns 400 if it's missing.\",\n ],\n [\n \"date\",\n \"No\",\n \"Query string\",\n \"Optional YYYY-MM-DD. Without it you get the recurring schedule and the days each flight operates.\",\n ],\n ],\n },\n {\n t: \"code\",\n v: `GET ${BASE}/flights?origin=BLR&destination=DXB\n\n{\n \"query\": { \"origin\": \"BLR\", \"destination\": \"DXB\", \"date\": null, \"weekday\": null },\n \"count\": 1,\n \"flights\": [\n {\n \"flight_number\": \"EA701\",\n \"origin\": \"Bengaluru (BLR)\",\n \"destination\": \"Dubai (DXB)\",\n \"departure_time\": \"02:30\",\n \"arrival_time\": \"04:45\",\n \"aircraft\": \"787-9\",\n \"days_of_week\": \"Daily\"\n }\n ]\n}`,\n },\n { t: \"h2\", v: \"POST /verifybooking\" },\n {\n t: \"p\",\n v: \"Checks a booking reference against a last name. On a match it writes a row into verify_tokens and returns that token.\",\n },\n {\n t: \"table\",\n head: [\"Parameter\", \"Required\", \"Sent as\", \"Notes\"],\n rows: [\n [\n \"booking_ref\",\n \"Yes\",\n \"Request body\",\n \"The reference the caller quotes. This is the real secret.\",\n ],\n [\n \"last_name\",\n \"Yes\",\n \"Request body\",\n \"Matched loosely: small typos and near-misses pass, because it arrives by voice.\",\n ],\n ],\n },\n {\n t: \"code\",\n v: `Request body: { \"booking_ref\": \"EA9M2X8\", \"last_name\": \"Harrington\" }\n\nResponse (match): { \"authenticated\": true, \"token\": \"b6e1f2b0-...\", \"flight_number\": \"EA701\", \"travel_date\": \"2026-09-11\" }\nResponse (no match): { \"authenticated\": false }`,\n },\n { t: \"h2\", v: \"GET /flightstatus\" },\n {\n t: \"p\",\n v: \"Returns the live status of the flight the token was issued for.\",\n },\n {\n t: \"table\",\n head: [\"Parameter\", \"Required\", \"Sent as\", \"Notes\"],\n rows: [\n [\n \"token\",\n \"Yes\",\n \"Query string\",\n \"The token from /verifybooking. Anything else returns 401.\",\n ],\n ],\n },\n {\n t: \"code\",\n v: `{\n \"flight_number\": \"EA701\",\n \"travel_date\": \"2026-09-11\",\n \"status\": \"Delayed\",\n \"delay_minutes\": 45,\n \"new_departure_time\": \"03:15\",\n \"gate\": \"B12\"\n}`,\n },\n { t: \"h2\", v: \"Why there's a token\" },\n {\n t: \"p\",\n v: 'A caller asking about \"my flight\" has to prove the booking is theirs before anything is read back. That check happens once, in /verifybooking.',\n },\n {\n t: \"p\",\n v: \"Rather than hand the booking reference back and trust it to be passed along correctly, the backend generates a random token, stores it alongside the flight and travel day it just resolved, and returns only the token. /flightstatus looks that row up and answers for that flight, so the second call carries no customer details at all.\",\n },\n ],\n },\n {\n id: \"challenge\",\n nav: \"The challenge\",\n group: \"Orientation\",\n title: \"The challenge\",\n blocks: [\n {\n t: \"goal\",\n lines: [\n \"Build one AIDA Realtime Voice agent for EnterpriseAir that can, in a single continuous phone call:\",\n \"1. Look up scheduled flights between two places and read them back naturally.\",\n \"2. Verify a caller's booking (reference + last name), then report live status: on time, delayed, or cancelled.\",\n \"3. Answer general policy questions (baggage, check-in, cancellations, fares, refunds, routes) from a knowledge base, without guessing.\",\n \"4. Decline anything else, such as booking a flight or issuing a refund, and offer a human agent.\",\n \"It opens every call with a welcome message naming all three capabilities, sounds natural aloud, and never reveals the tools or database behind it.\",\n ],\n },\n { t: \"h2\", v: \"Before you start\" },\n {\n t: \"p\",\n v: \"Three things to open now. Keep them in tabs alongside this page.\",\n },\n {\n t: \"download\",\n href: LINKS.aida,\n label: \"Log in to AIDA\",\n note: \"demo-chatbot2go.enterprisebot.co\",\n cta: \"Open\",\n },\n {\n t: \"download\",\n href: LINKS.docs,\n label: \"Documents for the knowledge base\",\n note: \"Four policy documents you'll upload in Task 1\",\n },\n {\n t: \"download\",\n href: LINKS.valtown,\n label: \"The EnterpriseAir backend on val.town\",\n note: \"The source behind the three endpoints\",\n cta: \"Open\",\n },\n { t: \"h2\", v: \"The names you'll create\" },\n {\n t: \"warn\",\n label: \"Everything you create is prefixed with your initials.\",\n text:\n \"Two people cannot create a tool of the same name. Tools are also shared between agents, so two different agents can call the same tool. That isn't what we want here: you want to call only the tools you created yourself. Every name below is already personalised, so use them exactly as written.\",\n },\n {\n t: \"table\",\n head: [\"What you create\", \"Your name for it\"],\n rows: [\n [\"Your agent\", \"xx_EnterpriseAir_Voice\"],\n [\"Your knowledge base\", \"xx_EnterpriseAir_Policies\"],\n [\n \"Your tools\",\n \"xx_enterpriseair_getflights, xx_enterpriseair_verifybooking, xx_enterpriseair_getstatus\",\n ],\n ],\n },\n ],\n },\n\n // ===================== TASK 1 =====================\n {\n id: \"task1\",\n nav: \"Build the knowledge base\",\n group: \"Tasks\",\n taskNo: 1,\n title: \"Task 1 — Build the knowledge base\",\n blocks: [\n {\n t: \"task\",\n label: \"DocBrain\",\n points: 100,\n lines: [\n \"Indexing takes longer than anything else in this lab, so it goes first and runs while you work.\",\n \"Create a knowledge base in DocBrain, upload EnterpriseAir's four policy documents, and start indexing.\",\n ],\n },\n {\n t: \"p\",\n v: \"The four documents are Conditions of Carriage, Baggage Policy, FAQ, and Route Network & Destinations.\",\n },\n {\n t: \"download\",\n href: LINKS.docs,\n label: \"EnterpriseAir policy documents\",\n note: \"SharePoint folder · 4 files\",\n },\n {\n t: \"callout\",\n label: \"Don't wait for indexing to finish.\",\n text: \"Move on to Task 2 and let it run in the background.\",\n },\n {\n t: \"hint\",\n id: \"t1-nudge\",\n kind: \"nudge\",\n cost: 10,\n label: \"Nudge\",\n lines: [\n \"DocBrain has its own hub, reached from the horizontal menu at the top right, separate from the agent. Build the knowledge base there now; you'll attach it to the agent in Task 6.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t1-steps\",\n kind: \"steps\",\n cost: 25,\n label: \"Step-by-step\",\n lines: [\n \"1. Open the DocBrain hub from the horizontal menu at the top right.\",\n \"2. Create a new knowledge base.\",\n \"3. Name it xx_EnterpriseAir_Policies. The prefix keeps your knowledge base separate from everyone else's.\",\n \"4. Upload all four documents.\",\n \"5. Start indexing and move straight on to Task 2. Don't wait for it to finish.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t1-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [\n \"Name: xx_EnterpriseAir_Policies\",\n \"Documents: Conditions of Carriage, Baggage Policy, FAQ, Route Network & Destinations.\",\n \"Leave it indexing and start Task 2.\",\n ],\n guide: {\n href: LINKS.guideKb,\n label: \"Building the knowledge base (PDF walkthrough)\",\n },\n },\n ],\n },\n\n // ===================== TASK 2 =====================\n {\n id: \"task2\",\n nav: \"Workbook: design the agent\",\n group: \"Tasks\",\n taskNo: 2,\n title: \"Task 2 — The workbook: design the agent\",\n blocks: [\n {\n t: \"p\",\n v: \"This task happens on paper, before you touch AIDA. You'll finish it holding the agent's job list, its tool plan and its three prompts, ready to enter in Tasks 3 to 6.\",\n },\n {\n t: \"download\",\n href: LINKS.workbook,\n label: \"LabWorkBook.docx\",\n note:\n \"Save it to your machine and fill it in as you work through this task\",\n },\n {\n t: \"p\",\n v: \"Work through the three parts below. Each table here mirrors one in the workbook, and every hint fills in a row so you can check yourself as you go.\",\n },\n {\n t: \"task\",\n label: \"Plan before you build\",\n points: 100,\n lines: [\n \"Work through the three parts below, filling in the workbook as you go.\",\n \"Everything you write here gets entered into AIDA later, so write it as you'd want to paste it.\",\n ],\n },\n\n { t: \"h2\", v: \"Part 1 — Define the jobs\" },\n {\n t: \"p\",\n v: \"Our agent has a list of jobs it needs to do. Let's define each job by naming the job, the API the agent will use to obtain the information, and the boundary the agent should not cross.\",\n },\n {\n t: \"p\",\n v: \"EnterpriseAir's four jobs are named for you. Fill in the rest in your workbook. The first row is done as an example; open a hint to fill in any other row.\",\n },\n {\n t: \"htable\",\n head: [\"Job\", \"Source of truth\", \"Boundary\"],\n rows: [\n {\n cells: [\n \"Schedule lookup\",\n \"GET /flights\",\n \"None, the timetable is public.\",\n ],\n },\n {\n cells: [\n \"Flight status\",\n \"POST /verifybooking, then GET /flightstatus\",\n \"Verify before saying anything about the flight.\",\n ],\n hint: { id: \"t2-job-status\", cost: 10 },\n },\n {\n cells: [\n \"Policy question\",\n \"The xx_EnterpriseAir_Policies knowledge base\",\n \"Hand over to a human if the knowledge base doesn't cover it.\",\n ],\n hint: { id: \"t2-job-policy\", cost: 10 },\n },\n {\n cells: [\n \"Anything else\",\n \"None\",\n \"Decline and offer a human agent.\",\n ],\n hint: { id: \"t2-job-else\", cost: 10 },\n },\n ],\n },\n\n { t: \"h2\", v: \"Part 2 — Define the tools\" },\n {\n t: \"p\",\n v: \"Tools and agents are the components of agentic AI. An agentic tool is a software component that empowers an AI agent to interact with its external environment and perform specific, pre-defined actions. It is the bridge between a large language model's reasoning and the real world.\",\n },\n {\n t: \"ul\",\n v: [\n \"An API tool calls an external system. It needs a method, a URL, parameters, and a description.\",\n \"A KB tool retrieves from a DocBrain knowledge base. It needs a description only.\",\n \"The description is the only thing the model sees when deciding whether to call a tool. Write it as an instruction, and always say when not to call it.\",\n \"Every parameter is marked required or optional, and marked as a query parameter or not. Follow the backend: a GET reads its values from the query string, a POST reads them from the body.\",\n ],\n },\n { t: \"h3\", v: \"Two of your tools are a chain\" },\n { t: \"p\", v: \"Checking a booking takes two tools, run in order.\" },\n {\n t: \"ul\",\n v: [\n \"xx_enterpriseair_verifybooking takes the booking reference and the last name, and returns a token.\",\n \"xx_enterpriseair_getstatus takes that token as its parameter, and returns the flight's status.\",\n ],\n },\n {\n t: \"p\",\n v: \"The second tool never sees the booking reference or the name. It only receives the token the first tool issued, which is what makes the check trustworthy.\",\n },\n {\n t: \"p\",\n v: \"Four tools cover the whole agent. Fill in the rest in your workbook, using the backend reference. The first row is done as an example; open a hint to fill in any other row.\",\n },\n {\n t: \"htable\",\n head: [\n \"Tool name\",\n \"Type\",\n \"Method\",\n \"Endpoint\",\n \"Parameters (required / query)\",\n ],\n rows: [\n {\n cells: [\n \"xx_enterpriseair_getflights\",\n \"API\",\n \"GET\",\n \"/flights\",\n \"origin (yes / query), destination (yes / query)\",\n ],\n },\n {\n cells: [\n \"xx_enterpriseair_verifybooking\",\n \"API\",\n \"POST\",\n \"/verifybooking\",\n \"booking_ref (yes / body), last_name (yes / body)\",\n ],\n hint: { id: \"t2-tool-verify\", cost: 10 },\n },\n {\n cells: [\n \"xx_enterpriseair_getstatus\",\n \"API\",\n \"GET\",\n \"/flightstatus\",\n \"token (yes / query)\",\n ],\n hint: { id: \"t2-tool-status\", cost: 10 },\n },\n {\n cells: [\"xx_EnterpriseAir_Policies\", \"KB\", \"-\", \"-\", \"None\"],\n hint: { id: \"t2-tool-kb\", cost: 10 },\n },\n ],\n },\n {\n t: \"p\",\n v: \"A POST carries its values in the body, so nothing on verifybooking is a query parameter. A GET carries its values in the URL, so the token on getstatus must be.\",\n },\n { t: \"h3\", v: \"Write each tool's description\" },\n {\n t: \"p\",\n v: \"One line on when to call it, one on when not to. This is the text the model reads, so write it as an instruction to the model.\",\n },\n {\n t: \"hint\",\n id: \"t2-desc-flights\",\n kind: \"scaffold\",\n cost: 10,\n label: \"xx_enterpriseair_getflights description\",\n code: DESC_GETFLIGHTS,\n },\n {\n t: \"hint\",\n id: \"t2-desc-verify\",\n kind: \"scaffold\",\n cost: 10,\n label: \"xx_enterpriseair_verifybooking description\",\n code: DESC_VERIFYBOOKING,\n },\n {\n t: \"hint\",\n id: \"t2-desc-status\",\n kind: \"scaffold\",\n cost: 10,\n label: \"xx_enterpriseair_getstatus description\",\n code: DESC_GETSTATUS,\n },\n {\n t: \"hint\",\n id: \"t2-desc-kb\",\n kind: \"scaffold\",\n cost: 10,\n label: \"xx_EnterpriseAir_Policies description\",\n code: DESC_KB,\n },\n\n { t: \"h2\", v: \"Part 3 — Write the prompts\" },\n { t: \"p\", v: \"AIDA splits the agent's brain across three fields.\" },\n {\n t: \"table\",\n head: [\"Field\", \"What it controls\"],\n rows: [\n [\n \"Welcome message\",\n \"The first thing the caller hears: who the agent is, what it can do, an invitation to speak.\",\n ],\n [\n \"Instructions\",\n \"The job list, how to recognise each job, which tool to call in which order, and the refusal boundary.\",\n ],\n [\n \"Response guidelines\",\n \"The agent's response style: length, phrasing, how numbers, codes and times are spoken.\",\n ],\n ],\n },\n { t: \"h3\", v: \"Using Claude or ChatGPT to draft these\" },\n {\n t: \"p\",\n v: \"Drafting prompts is a good use of an AI assistant, as long as you give it something to copy. The technique is called few-shot prompting: show the model two or three examples of the output you want, then ask for the next one. Examples move a model further than description does.\",\n },\n {\n t: \"example\",\n label: \"Worked example — a few-shot prompt\",\n lines: [\n '\"Here are two welcome messages for voice agents:',\n \"1. Welcome to Harbour Hotels. I'm your virtual assistant. I can check a reservation, tell you about our facilities, or answer questions about your stay. How can I help you today?\",\n \"2. Welcome to Meridian Bank. I'm your virtual assistant. I can check a balance, explain a transaction, or answer questions about our accounts. How can I help you today?\",\n \"Write one in the same style for EnterpriseAir, an airline. It can check flight schedules, look up a booking's status, and answer policy questions.\\\"\",\n \"Then check the output yourself: right length, right capabilities, nothing invented.\",\n ],\n },\n { t: \"h3\", v: \"The welcome message\" },\n {\n t: \"p\",\n v: \"Name the brand, state the three capabilities, invite the caller to speak. Nothing else.\",\n },\n {\n t: \"hint\",\n id: \"t2-welcome\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Welcome message\",\n code: WELCOME_MSG,\n },\n { t: \"h3\", v: \"The instructions\" },\n {\n t: \"p\",\n v: \"You are writing the instructions prompt. It is made of five parts, written one after the other in this order.\",\n },\n {\n t: \"table\",\n head: [\"Part\", \"What it contains\"],\n rows: [\n [\"Role\", \"Who the agent is and the three jobs it covers.\"],\n [\n \"Flight schedule inquiries\",\n \"How to recognise one, which tool to call, what to read back.\",\n ],\n [\n \"Flight status inquiries\",\n \"What to ask for and in what order, both tools in sequence, and what to do when verification fails.\",\n ],\n [\n \"General policy questions\",\n \"Answer from the knowledge base only, and what to do when it isn't covered.\",\n ],\n [\n \"Hard rules and output format\",\n \"What is never allowed in any job, and how answers should be worded.\",\n ],\n ],\n },\n {\n t: \"p\",\n v: \"Write one part at a time. Each hint below reveals that part as it goes into AIDA.\",\n },\n {\n t: \"hint\",\n id: \"t2-ins-role\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Role\",\n code: INS_ROLE,\n },\n {\n t: \"hint\",\n id: \"t2-ins-sched\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Block 1 — schedule questions\",\n code: INS_SCHEDULE,\n },\n {\n t: \"hint\",\n id: \"t2-ins-status\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Block 2 — booking status\",\n code: INS_STATUS,\n },\n {\n t: \"hint\",\n id: \"t2-ins-policy\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Block 3 — policy questions\",\n code: INS_POLICY,\n },\n {\n t: \"hint\",\n id: \"t2-ins-rules\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Hard rules and output format\",\n code: INS_RULES,\n },\n {\n t: \"callout\",\n label: \"That's the whole prompt.\",\n text:\n \"Those five parts, one after the other, are the single piece of text you paste into the Instructions field in Task 3.\",\n },\n {\n t: \"warn\",\n label: \"Watch out for this trap:\",\n text:\n 'A hard rule like \"decline anything about refunds\" will contradict the knowledge base, which already covers refund policy, the moment a customer asks how refunds work. Draw the line by action vs. information, not by topic: explaining a policy is always fine if the knowledge base covers it; carrying out an action (booking, refunding, cancelling) is always declined.',\n },\n { t: \"h3\", v: \"The response guidelines\" },\n {\n t: \"p\",\n v: \"This prompt defines the agent's response style. Cover at least: response length, spoken language and punctuation, tone, asking one question at a time, how flight numbers and gate letters are pronounced, how times are spoken, and the order to state a delay in.\",\n },\n {\n t: \"hint\",\n id: \"t2-voice\",\n kind: \"scaffold\",\n cost: 10,\n label: \"Response guidelines\",\n code: VOICE_RULES,\n },\n { t: \"h2\", v: \"Check your plan\" },\n {\n t: \"ul\",\n v: [\n \"Every job has a named source of truth, or an explicit refusal.\",\n \"Every tool description says when to call it and when not to.\",\n \"Nothing in your hard rules contradicts a topic the knowledge base covers.\",\n \"The welcome message names every capability the agent actually has.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t2-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [\n \"The completed workbook, with every table filled and all three prompts written out, is linked below.\",\n ],\n guide: { href: LINKS.workbookDone, label: \"Completed workbook (PDF)\" },\n },\n ],\n },\n\n // ===================== TASK 3 =====================\n {\n id: \"task3\",\n nav: \"Create the agent and its prompts\",\n group: \"Tasks\",\n taskNo: 3,\n title: \"Task 3 — Create the agent and its prompts\",\n blocks: [\n {\n t: \"task\",\n label: \"The agent shell and all three prompts\",\n points: 100,\n lines: [\n \"Everything you wrote in the workbook now goes into AIDA.\",\n \"Create a Realtime Voice agent, then enter its identity, welcome message, instructions and response guidelines.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t3-nudge\",\n kind: \"nudge\",\n cost: 10,\n label: \"Nudge\",\n lines: [\n \"AIDA lets you create Agents as well as Bots. Make sure you are creating an Agent, then pick the voice type on the Agent Mode screen.\",\n \"The welcome message sits with the company info. The instructions and the response guidelines sit under Prompts.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t3-steps\",\n kind: \"steps\",\n cost: 25,\n label: \"Step-by-step\",\n lines: [\n \"1. From the Assistants page, select Create Assistant, then Agent.\",\n \"2. On Agent Mode, select Realtime Voice, then Continue.\",\n \"3. Fill in the agent's name, primary language and organization name. It is ok to leave the website empty for this build.\",\n \"4. Paste the welcome message from your workbook.\",\n '5. Find the field called \"What should your AI assistant do?\" and paste in the instructions you wrote. The tools won\\'t exist yet; build them under exactly these names in Tasks 4 and 5.',\n \"6. In Prompts → Response Guidelines, paste your response guidelines.\",\n ],\n },\n {\n t: \"callout\",\n label: \"About the prompts in the solution:\",\n text:\n \"What goes into AIDA is what you wrote in your workbook. The versions below are ours, for comparison. Use them if you're stuck, or check yours against them and keep your own wording where it works.\",\n },\n {\n t: \"hint\",\n id: \"t3-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [\"Field values:\"],\n table: {\n head: [\"Field\", \"What to enter\"],\n rows: [\n [\"AI Assistant Name\", \"xx_EnterpriseAir_Voice\"],\n [\"Primary Language\", \"English\"],\n [\"Organization Name\", \"EnterpriseAir\"],\n [\"Website URL\", \"Leave blank for this build.\"],\n ],\n },\n codes: [\n { label: \"Welcome message\", v: WELCOME_MSG },\n { label: \"Instructions\", v: INSTRUCTIONS_FULL },\n { label: \"Response guidelines\", v: VOICE_RULES },\n ],\n guide: {\n href: LINKS.guideAgent,\n label: \"Creating the agent and its prompts (PDF walkthrough)\",\n },\n },\n ],\n },\n\n // ===================== TASK 4 =====================\n {\n id: \"task4\",\n nav: \"Build the schedule tool\",\n group: \"Tasks\",\n taskNo: 4,\n title: \"Task 4 — Build the schedule-lookup tool\",\n blocks: [\n {\n t: \"task\",\n label: \"xx_enterpriseair_getflights\",\n points: 100,\n lines: [\n \"This tool looks up which flights run between two places, and what time they depart and arrive.\",\n \"Build it from your workbook plan and assign it to the agent.\",\n ],\n },\n {\n t: \"callout\",\n label: \"What to work from:\",\n text:\n \"The GET /flights reference in The backend section has the parameters and the response. Your workbook has the tool name, the method and the description you drafted.\",\n },\n {\n t: \"hint\",\n id: \"t4-nudge\",\n kind: \"nudge\",\n cost: 10,\n label: \"Nudge\",\n lines: [\n \"Simplest of the three tools: one call, no verification - a published timetable isn't sensitive. Check the GET /flights reference for the parameters it expects.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t4-steps\",\n kind: \"steps\",\n cost: 25,\n label: \"Step-by-step\",\n lines: [\n \"1. Go to Tools, select Create, and choose API Tool.\",\n \"2. Set the method to GET and the URL to the /flights endpoint on the base URL.\",\n \"3. Add origin as a String parameter. Mark it Required, because the backend errors without it. Mark Query Parameter On, because this is a GET and the backend reads it from the query string.\",\n \"4. Add destination the same way: String, Required, Query Parameter On.\",\n \"5. Enter the description you drafted: when to use it, and when not to.\",\n \"6. Assign the tool to your agent - a tool that exists but isn't assigned is never called.\",\n ],\n },\n {\n t: \"example\",\n label: \"Why required, and why query\",\n lines: [\n \"Required: the backend checks for origin and destination and returns a 400 error if either is missing, so the model has to supply both every time.\",\n \"Query Parameter On: a GET request has no body. The backend reads both values off the URL, as /flights?origin=BLR&destination=DXB. With Query Parameter Off they would never arrive.\",\n 'The optional date parameter is left out of this build on purpose: without it the endpoint returns the recurring schedule, which is what a caller asking \"what flies to Dubai\" actually wants.',\n ],\n },\n {\n t: \"hint\",\n id: \"t4-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [\"Tool name: xx_enterpriseair_getflights. Field values:\"],\n table: {\n head: [\"Field\", \"Value\"],\n rows: [\n [\"Tool Type\", \"API Tool\"],\n [\"Tool Name\", \"xx_enterpriseair_getflights\"],\n [\"Method\", \"GET\"],\n [\"URL\", BASE + \"/flights\"],\n [\"Response Type\", \"JSON\"],\n [\"Time out\", \"60\"],\n ],\n },\n codes: [{ label: \"Description\", v: DESC_GETFLIGHTS }],\n table2: {\n head: [\"Parameter\", \"Type\", \"Required\", \"Query Param\", \"Description\"],\n rows: [\n [\n \"origin\",\n \"String\",\n \"Yes\",\n \"On\",\n \"Airport code or city name the customer is flying from\",\n ],\n [\n \"destination\",\n \"String\",\n \"Yes\",\n \"On\",\n \"Airport code or city name the customer is flying to\",\n ],\n ],\n },\n guide: {\n href: LINKS.guideFlights,\n label: \"Building the schedule-lookup tool (PDF walkthrough)\",\n },\n },\n ],\n },\n\n // ===================== TASK 5 =====================\n {\n id: \"task5\",\n nav: \"Build the verify-then-fetch chain\",\n group: \"Tasks\",\n taskNo: 5,\n title: \"Task 5 — Build the verify-then-fetch chain\",\n blocks: [\n {\n t: \"task\",\n label: \"xx_enterpriseair_verifybooking + xx_enterpriseair_getstatus\",\n points: 100,\n lines: [\n \"These two tools work as a pair. The first checks a booking reference against a last name and returns a token; the second turns that token into the flight's status.\",\n \"Build both and assign them to the agent.\",\n ],\n },\n {\n t: \"callout\",\n label: \"What to work from:\",\n text:\n \"The POST /verifybooking and GET /flightstatus references in The backend section have the parameters and the responses. Your workbook has both tool names and the descriptions you drafted.\",\n },\n {\n t: \"hint\",\n id: \"t5-nudge\",\n kind: \"nudge\",\n cost: 10,\n label: \"Nudge\",\n lines: [\n \"The POST tool turns booking reference + last name into a token; the GET tool turns that token into status data. The second tool never sees the booking reference, only the token.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t5-steps\",\n kind: \"steps\",\n cost: 25,\n label: \"Step-by-step\",\n lines: [\n \"1. Create xx_enterpriseair_verifybooking: POST to /verifybooking. Two required String parameters, booking_ref and last_name, Query Param Off - they travel in the body.\",\n \"2. Create xx_enterpriseair_getstatus: GET to /flightstatus. One required String parameter, token, Query Param On - it must arrive as ?token=... in the URL.\",\n \"3. Describe verifybooking: call it once both values are known; the last name doesn't need an exact match.\",\n \"4. Describe getstatus: call it only after verifybooking returns authenticated true, passing its token.\",\n \"5. Assign both tools to the agent.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t5-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [\"Tool 1 of 2 — xx_enterpriseair_verifybooking. Field values:\"],\n table: {\n head: [\"Field\", \"Value\"],\n rows: [\n [\"Tool Type\", \"API Tool\"],\n [\"Tool Name\", \"xx_enterpriseair_verifybooking\"],\n [\"Method\", \"POST\"],\n [\"URL\", BASE + \"/verifybooking\"],\n [\"Response Type\", \"JSON\"],\n [\"Time out\", \"60\"],\n ],\n },\n codes: [{ label: \"Description\", v: DESC_VERIFYBOOKING }],\n table2: {\n head: [\"Parameter\", \"Type\", \"Required\", \"Query Param\", \"Description\"],\n rows: [\n [\n \"booking_ref\",\n \"String\",\n \"Yes\",\n \"Off\",\n \"The customer's booking reference\",\n ],\n [\n \"last_name\",\n \"String\",\n \"Yes\",\n \"Off\",\n \"The last name on the booking (fuzzy-matched)\",\n ],\n ],\n },\n divider2: \"Tool 2 of 2 — xx_enterpriseair_getstatus. Field values:\",\n table3: {\n head: [\"Field\", \"Value\"],\n rows: [\n [\"Tool Type\", \"API Tool\"],\n [\"Tool Name\", \"xx_enterpriseair_getstatus\"],\n [\"Method\", \"GET\"],\n [\"URL\", BASE + \"/flightstatus\"],\n [\"Response Type\", \"JSON\"],\n [\"Time out\", \"60\"],\n ],\n },\n codes2: [{ label: \"Description\", v: DESC_GETSTATUS }],\n table4: {\n head: [\"Parameter\", \"Type\", \"Required\", \"Query Param\", \"Description\"],\n rows: [\n [\n \"token\",\n \"String\",\n \"Yes\",\n \"On\",\n \"The token returned by xx_enterpriseair_verifybooking\",\n ],\n ],\n },\n guide: {\n href: LINKS.guideChain,\n label: \"Building the verify-then-fetch chain (PDF walkthrough)\",\n },\n },\n {\n t: \"callout\",\n label: \"Now that the tools exist:\",\n text:\n \"Go back to Prompts → Instructions and check the tool names you entered in Task 3 match the tools you just built, character for character. A name that doesn't match is a tool that never gets called.\",\n },\n ],\n },\n\n // ===================== TASK 6 =====================\n {\n id: \"task6\",\n nav: \"Attach the knowledge base and publish\",\n group: \"Tasks\",\n taskNo: 6,\n title: \"Task 6 — Attach the knowledge base and publish\",\n blocks: [\n {\n t: \"task\",\n label: \"xx_EnterpriseAir_Policies\",\n points: 100,\n lines: [\n \"The knowledge base you started in Task 1 should be indexed by now.\",\n \"Attach it to the agent as a KB tool, describe when to use it, and publish.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t6-nudge\",\n kind: \"nudge\",\n cost: 10,\n label: \"Nudge\",\n lines: [\n \"A KB tool isn't an API Tool - look for a dedicated knowledge-base tool type on the Tools page.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t6-steps\",\n kind: \"steps\",\n cost: 25,\n label: \"Step-by-step\",\n lines: [\n \"1. In the DocBrain hub, confirm xx_EnterpriseAir_Policies shows as fully indexed.\",\n \"2. On the agent's Tools page, create a KB tool pointing at xx_EnterpriseAir_Policies.\",\n \"3. Assign it to the agent.\",\n \"4. Describe when to use it (policy, baggage, routes, check-in) and when not to (not for live flight status).\",\n \"5. Select Publish. Nothing you've configured takes effect until you publish.\",\n ],\n },\n {\n t: \"hint\",\n id: \"t6-sol\",\n kind: \"solution\",\n cost: 50,\n label: \"Full solution\",\n lines: [],\n codes: [{ label: \"Description\", v: DESC_KB }],\n guide: {\n href: LINKS.guidePublish,\n label:\n \"Attaching the knowledge base and publishing (PDF walkthrough)\",\n },\n },\n ],\n },\n\n // ===================== TESTING =====================\n {\n id: \"testing\",\n nav: \"Testing script\",\n group: \"Verification\",\n title: \"Testing script\",\n blocks: [\n { t: \"h2\", v: \"How to run a test\" },\n {\n t: \"p\",\n v: \"Click Publish first, then Try it. You can now talk to your agent.\",\n },\n {\n t: \"p\",\n v: \"Each working scenario is worth 10 bonus points, up to 70 total. Tick them off as you go.\",\n },\n {\n t: \"test\",\n id: \"v1\",\n label: \"Schedule lookup\",\n text:\n 'Say: \"Are there any flights from Bengaluru to Dubai?\" Expect: EA701, Daily, 02:30 to 04:45.',\n },\n {\n t: \"test\",\n id: \"v2\",\n label: \"Delay check — happy path, Delayed\",\n text:\n \"Give booking EA9M2X8, last name Harrington. Expect: EA701, Delayed 45 minutes, new departure 03:15, gate B12.\",\n },\n {\n t: \"test\",\n id: \"v3\",\n label: \"Delay check — happy path, Cancelled\",\n text:\n \"Give booking EA7R3W9, last name Mehta. Expect: EA731, Cancelled.\",\n },\n {\n t: \"test\",\n id: \"v4\",\n label: \"Delay check — near-miss last name\",\n text:\n 'Give booking EA7R3W9 with a mistranscribed last name, like \"Meta\" instead of \"Mehta.\" Expect: this still succeeds - the match is fuzzy on purpose.',\n },\n {\n t: \"test\",\n id: \"v5\",\n label: \"Delay check — wrong last name\",\n text:\n 'Give booking EA2K8L1 with an unrelated last name, like \"Smith.\" Expect: the agent asks you to spell the name, tries once more with the same booking reference, then fails gracefully with a polite \"details didn\\'t match\", without saying which field was wrong.',\n },\n {\n t: \"test\",\n id: \"v6\",\n label: \"Policy question via DocBrain\",\n text:\n 'Ask: \"What\\'s my checked baggage allowance in Economy on a domestic flight?\" Expect an answer sourced from the Baggage Policy doc.',\n },\n {\n t: \"test\",\n id: \"v7\",\n label: \"Out of scope\",\n text:\n \"Ask it to book a new flight: expect a graceful decline and a human-agent offer. Separately, confirm a refund-policy question (not a refund request) is still answered from the knowledge base.\",\n },\n ],\n },\n {\n id: \"next\",\n nav: \"What's next\",\n group: \"Verification\",\n title: \"What's next\",\n blocks: [\n {\n t: \"p\",\n v: \"Congratulations. You've built a multi-capability AIDA voice agent end to end: knowledge base, tools, prompts and tests.\",\n },\n { t: \"scoreboard\" },\n ],\n },\n];",[],[191],["D",1789493347078],["D",1789494690166],{"_11":12,"_143":68},{"_11":12,"_138":139,"_67":68,"_30":-5,"_140":141,"_142":24},{"_40":129,"_13":16,"_34":37,"_130":91,"_131":-5,"_132":-5,"_133":-5,"_134":-5,"_135":-5,"_136":14,"_137":12},"ctx",{"_52":-5,"_162":236,"_237":238,"_239":240,"_241":24,"_242":24,"_243":24,"_244":103,"_245":-5,"_246":-7,"_247":-7,"_248":26,"_249":24,"_250":24,"_251":-5,"_252":24,"_253":24,"_254":24,"_255":-5},"ENV",{"_207":208,"_209":210,"_211":16,"_212":213,"_214":215,"_216":217,"_218":219,"_220":221,"_222":223,"_224":225,"_226":227,"_228":229,"_230":231,"_232":233,"_234":235},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

History · main.ts