[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_253":254,"_255":256},"routes/_app.x.$handle.$name",{"_101":100,"_33":46,"_110":91,"_27":111,"_13":14,"_112":-5,"_11":113,"_114":115,"_116":117,"_118":119,"_120":121,"_20":-5,"_122":-5,"_123":-5,"_124":25,"_125":126,"_127":25,"_128":-5,"_129":-5,"_130":131},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_44":45,"_33":46,"_39":47,"_48":-5,"_49":25,"_50":51,"_52":51,"_53":54,"_55":-5,"_18":56,"_57":58,"_59":60,"_61":-5,"_62":-5,"_63":64,"_65":66,"_67":68,"_69":70,"_71":72,"_73":-5,"_74":75,"_76":75,"_77":78,"_79":78,"_80":78,"_81":75,"_82":75,"_83":84,"_85":86,"_87":88,"_89":25},"version",18,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":30},"file",{"_31":32,"_33":28,"_34":35,"_36":38,"_39":40,"_41":14,"_42":43,"_24":25},"previous",{"_31":32,"_33":28,"_34":35,"_36":37},"deleted",false,"deletedHere","path","new-file-2883.ts","latestPath","tools.ts","fileId","257d9160-ad1b-11f1-96e7-1607ee4eb77e","name","fileType","script","content","// Step 2 of 4 — TOOLS. Two things the agent can *do*, both keyless Oslo open data.\n// Pure fetch, no Val.town imports, so you can run `node --test` on this file locally.\n\nconst CLIENT = \"oslovibecoding-first-agent\"; // both APIs ask you to identify your app\n\n// ---------- Entur: live departures from any Oslo stop ----------\nexport async function osloDepartures(stopName: string): Promise {\n const geo = await fetch(\n `https://api.entur.io/geocoder/v1/autocomplete?text=${\n encodeURIComponent(stopName)\n }&layers=venue&size=1&boundary.county_ids=03`,\n { headers: { \"ET-Client-Name\": CLIENT } },\n ).then((r) => r.json());\n const hit = geo.features?.[0]?.properties;\n if (!hit) return `No stop found for \"${stopName}\".`;\n\n const query = `query($id:String!){ stopPlace(id:$id){ name\n estimatedCalls(numberOfDepartures:8, timeRange:3600){ expectedDepartureTime realtime\n destinationDisplay{ frontText } serviceJourney{ line{ publicCode transportMode } } } } }`;\n const data = await fetch(\"https://api.entur.io/journey-planner/v3/graphql\", {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\", \"ET-Client-Name\": CLIENT },\n body: JSON.stringify({ query, variables: { id: hit.id } }),\n }).then((r) => r.json());\n const calls = data.data?.stopPlace?.estimatedCalls ?? [];\n if (calls.length === 0) {\n return `No departures in the next hour from ${hit.name}.`;\n }\n const lines = calls.map((c: any) => {\n const t = new Date(c.expectedDepartureTime).toLocaleTimeString(\"en-GB\", {\n timeZone: \"Europe/Oslo\",\n hour: \"2-digit\",\n minute: \"2-digit\",\n });\n return `${t} ${c.serviceJourney.line.transportMode} ${c.serviceJourney.line.publicCode} to ${c.destinationDisplay.frontText}`;\n });\n return `Departures from ${hit.name} (Oslo time):\\n${lines.join(\"\\n\")}`;\n}\n\n// ---------- Yr (MET Norway): the next few hours of weather in Oslo ----------\nexport async function osloWeather(hours = 6): Promise {\n const data = await fetch(\n \"https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=59.9111&lon=10.7528\",\n { headers: { \"User-Agent\": `${CLIENT} oslovibecoding.tech` } },\n ).then((r) => r.json());\n const series = (data.properties?.timeseries ?? []).slice(\n 0,\n Math.max(1, Math.min(hours, 12)),\n );\n if (series.length === 0) return \"Weather service returned nothing.\";\n const rows = series.map((s: any) => {\n const t = new Date(s.time).toLocaleTimeString(\"en-GB\", {\n timeZone: \"Europe/Oslo\",\n hour: \"2-digit\",\n minute: \"2-digit\",\n });\n const temp = s.data.instant.details.air_temperature;\n const next = s.data.next_1_hours;\n const rain = next?.details?.precipitation_amount ?? 0;\n const sky = next?.summary?.symbol_code ?? \"unknown\";\n return `${t}: ${temp}°C, ${rain} mm rain, ${sky}`;\n });\n return `Oslo weather, next ${rows.length} hours:\\n${rows.join(\"\\n\")}`;\n}\n\n// ---------- What the model sees: a name, a description, and the arguments ----------\nexport const toolSchemas = [\n {\n type: \"function\",\n function: {\n name: \"oslo_departures\",\n description:\n \"Live public transport departures in the next hour from a named stop in Oslo (tram, metro, bus, train).\",\n parameters: {\n type: \"object\",\n properties: {\n stop_name: {\n type: \"string\",\n description: \"Stop name, e.g. Jernbanetorget\",\n },\n },\n required: [\"stop_name\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"oslo_weather\",\n description:\n \"Hourly weather forecast for central Oslo: temperature, rain in mm, sky.\",\n parameters: {\n type: \"object\",\n properties: {\n hours: {\n type: \"integer\",\n description: \"How many hours ahead, 1 to 12\",\n default: 6,\n },\n },\n },\n },\n },\n] as const;\n\n// Map tool name -> function. The agent loop calls this when the model asks for a tool.\nexport const openDataTools: Record Promise> = {\n oslo_departures: (a) => osloDepartures(a.stop_name),\n oslo_weather: (a) => osloWeather(a.hours),\n};","// Step 2 of 4 — TOOLS. Two things the agent can *do*, both keyless Oslo open data.\n// Pure fetch, no Val.town imports, so you can run `node --test` on this file locally.\n\nconst CLIENT = \"oslovibecoding-first-agent\"; // both APIs ask you to identify your app\n\n// ---------- Entur: live departures from any Oslo stop ----------\nexport async function osloDepartures(stopName: string): Promise {\n const geo = await fetch(\n `https://api.entur.io/geocoder/v1/autocomplete?text=${\n encodeURIComponent(stopName)\n }&layers=venue&size=1&boundary.county_ids=03`,\n { headers: { \"ET-Client-Name\": CLIENT } },\n ).then((r) => r.json());\n const hit = geo.features?.[0]?.properties;\n if (!hit) return `No stop found for \"${stopName}\".`;\n\n const query = `query($id:String!){ stopPlace(id:$id){ name\n estimatedCalls(numberOfDepartures:8, timeRange:3600){ expectedDepartureTime realtime\n destinationDisplay{ frontText } serviceJourney{ line{ publicCode transportMode } } } } }`;\n const data = await fetch(\"https://api.entur.io/journey-planner/v3/graphql\", {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\", \"ET-Client-Name\": CLIENT },\n body: JSON.stringify({ query, variables: { id: hit.id } }),\n }).then((r) => r.json());\n const calls = data.data?.stopPlace?.estimatedCalls ?? [];\n if (calls.length === 0) {\n return `No departures in the next hour from ${hit.name}.`;\n }\n const lines = calls.map((c: any) => {\n const t = new Date(c.expectedDepartureTime).toLocaleTimeString(\"en-GB\", {\n timeZone: \"Europe/Oslo\",\n hour: \"2-digit\",\n minute: \"2-digit\",\n });\n return `${t} ${c.serviceJourney.line.transportMode} ${c.serviceJourney.line.publicCode} to ${c.destinationDisplay.frontText}`;\n });\n return `Departures from ${hit.name} (Oslo time):\\n${lines.join(\"\\n\")}`;\n}\n\n// ---------- Yr (MET Norway): the next few hours of weather in Oslo ----------\nexport async function osloWeather(hours = 6): Promise {\n const data = await fetch(\n \"https://api.met.no/weatherapi/locationforecast/2.0/compact?lat=59.9111&lon=10.7528\",\n { headers: { \"User-Agent\": `${CLIENT} oslovibecoding.tech` } },\n ).then((r) => r.json());\n const series = (data.properties?.timeseries ?? []).slice(\n 0,\n Math.max(1, Math.min(hours, 12)),\n );\n if (series.length === 0) return \"Weather service returned nothing.\";\n const rows = series.map((s: any) => {\n const t = new Date(s.time).toLocaleTimeString(\"en-GB\", {\n timeZone: \"Europe/Oslo\",\n hour: \"2-digit\",\n minute: \"2-digit\",\n });\n const temp = s.data.instant.details.air_temperature;\n const next = s.data.next_1_hours;\n const rain = next?.details?.precipitation_amount ?? 0;\n const sky = next?.summary?.symbol_code ?? \"unknown\";\n return `${t}: ${temp}°C, ${rain} mm rain, ${sky}`;\n });\n return `Oslo weather, next ${rows.length} hours:\\n${rows.join(\"\\n\")}`;\n}\n\n// ---------- What the model sees: a name, a description, and the arguments ----------\nexport const toolSchemas = [\n {\n type: \"function\",\n function: {\n name: \"oslo_departures\",\n description:\n \"Live public transport departures in the next hour from a named stop in Oslo (tram, metro, bus, train).\",\n parameters: {\n type: \"object\",\n properties: {\n stop_name: {\n type: \"string\",\n description: \"Stop name, e.g. Jernbanetorget\",\n },\n },\n required: [\"stop_name\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"oslo_weather\",\n description:\n \"Hourly weather forecast for central Oslo: temperature, rain in mm, sky.\",\n parameters: {\n type: \"object\",\n properties: {\n hours: {\n type: \"integer\",\n description: \"How many hours ahead, 1 to 12\",\n default: 6,\n },\n },\n },\n },\n },\n] as const;\n\n// Map tool name -> function. The agent loop calls this when the model asks for a tool.\nexport const openDataTools: Record Promise> = {\n oslo_departures: (a) => osloDepartures(a.stop_name),\n oslo_weather: (a) => osloWeather(a.hours),\n};\nmemory;","createdAt",["D",1789046847780],"branchVersion","branchId","bfef0492-d163-4418-ae3c-5faf4f1ecca3","id","f90ddc54-2774-419c-93d1-6657b655cba3","oslo-assistant",["D",1789046612776],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Oslo Vibe Coding S07: free Telegram assistant for Oslo","imageUrl","e5a6aed0-75b5-4515-9a1a-04c8a7847d1b","pinnedFileId","2f9824ee-ad1b-11f1-9bc4-1607ee4eb77e","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","01a08b7a-97a7-722a-b4e2-f98d180c34ba","tursoName","f90ddc542774419c93d16657b655cba3","tursoHostname","f90ddc542774419c93d16657b655cba3-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789488693443],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_101":100,"_109":105},"org",{"_101":100,"_102":103,"_104":105,"_18":56,"_106":107,"_108":25},"branch",{"_13":90,"_33":91,"_44":43,"_92":45,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":-5,"_98":46,"_99":100},"isOwned",39,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","akijain2000","handle","displayName","Akshat Jain","avatar","https://avatars.githubusercontent.com/u/42636196?v=4","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_44":45,"_33":46,"_39":248,"_48":-5,"_49":25,"_50":51,"_52":51,"_53":54,"_55":-5,"_18":56,"_57":58,"_59":60,"_61":-5,"_62":-5,"_63":64,"_65":66,"_67":68,"_69":70,"_71":249,"_73":-5,"_74":75,"_76":75,"_77":78,"_79":78,"_80":78,"_81":75,"_82":75,"_83":250,"_85":251,"_87":252,"_89":25},"tree",[211,212,213,214,215,216,217,218,219],"files",[211,212,213,214,215,216,217,218,219],"counts",{"_201":64,"_202":147,"_203":64,"_204":205,"_206":207},"contents",[132,133,134,135,136,137,138,139,140],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_44":193,"_33":194,"_31":195,"_143":-5,"_42":43,"_41":196,"_145":25,"_39":197,"_13":198,"_34":35,"_36":199,"_24":25,"_149":-5,"_27":200},{"_44":188,"_33":189,"_31":190,"_143":-5,"_42":43,"_41":147,"_145":25,"_39":191,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_27":192},{"_44":184,"_33":28,"_31":32,"_143":-5,"_42":43,"_41":14,"_145":25,"_39":185,"_13":186,"_34":35,"_36":38,"_24":25,"_149":-5,"_27":187},{"_44":177,"_33":178,"_31":179,"_143":-5,"_42":43,"_41":180,"_145":25,"_39":181,"_13":147,"_34":35,"_36":182,"_24":25,"_149":-5,"_27":183},{"_44":170,"_33":171,"_31":172,"_143":-5,"_42":43,"_41":173,"_145":25,"_39":174,"_13":147,"_34":35,"_36":175,"_24":25,"_149":-5,"_27":176},{"_44":164,"_33":165,"_31":166,"_143":-5,"_42":43,"_41":167,"_145":25,"_39":168,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_27":169},{"_44":157,"_33":158,"_31":159,"_143":-5,"_42":43,"_41":160,"_145":25,"_39":161,"_13":147,"_34":35,"_36":162,"_24":25,"_149":-5,"_27":163},{"_44":151,"_33":152,"_31":153,"_143":-5,"_42":43,"_41":154,"_145":25,"_39":155,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_27":156},{"_44":141,"_33":142,"_31":58,"_143":-5,"_42":43,"_41":144,"_145":25,"_39":146,"_13":147,"_34":35,"_36":148,"_24":25,"_149":-5,"_27":150},"89787a07-6117-4f00-ad34-2db0ce9880ac","new-file-8315.ts","parentId",15,"latest",["D",1789046774524],1,"// HTTP trigger. Telegram POSTs every message here. One request = one turn of the agent.\n// Setup: add TELEGRAM_BOT_TOKEN to environment variables, then open this val's URL with ?setup once (see README).\nimport { type Message, runAgent } from \"./agent.ts\";\nimport { openDataTools, toolSchemas } from \"./tools.ts\";\nimport { history, memorySchemas, memoryTools, saveMessage } from \"./memory.ts\";\nimport { callModel } from \"./model.ts\";\n\nconst TOKEN = Deno.env.get(\"TELEGRAM_BOT_TOKEN\");\nconst tg = (method: string) => `https://api.telegram.org/bot${TOKEN}/${method}`;\n\nexport async function sendTelegram(chatId: string, text: string) {\n await fetch(tg(\"sendMessage\"), {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ chat_id: chatId, text }),\n });\n}\n\nconst SYSTEM = (now: string) =>\n `You are a personal assistant for someone living in Oslo. Today is ${now} (Europe/Oslo).\nBe brief and concrete. Use tools when they help: live departures, weather, memory, reminders.\nWhen the user tells you something about themselves (home stop, habits), store it with remember.\nFor \"should I leave now?\" questions, check departures and weather, then give ONE recommendation and a time to leave.`;\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n const selfUrl = url.origin + url.pathname;\n\n if (!TOKEN) {\n return new Response(\n \"Missing TELEGRAM_BOT_TOKEN. Add it under Environment variables, then reload.\",\n { status: 500 },\n );\n }\n\n // One-time setup: open this val's URL with ?setup in a browser. The val tells Telegram\n // \"send every message for my bot to this URL\". No token ever leaves the server.\n if (req.method === \"GET\" && url.searchParams.has(\"setup\")) {\n const result = await fetch(\n `${tg(\"setWebhook\")}?url=${encodeURIComponent(selfUrl)}`,\n ).then((r) => r.json());\n return Response.json({ webhook: selfUrl, telegram: result }, {\n status: result.ok ? 200 : 500,\n });\n }\n // Debugging: ?status shows what Telegram thinks the webhook is and any recent delivery error.\n if (req.method === \"GET\" && url.searchParams.has(\"status\")) {\n return Response.json(\n await fetch(tg(\"getWebhookInfo\")).then((r) => r.json()),\n );\n }\n if (req.method !== \"POST\") {\n return new Response(\n \"Oslo assistant is running.\\nOpen this URL with ?setup once to connect Telegram, then talk to the bot in Telegram.\\nOpen it with ?status to check the connection.\",\n );\n }\n const update = await req.json();\n const msg = update.message ?? update.edited_message;\n if (!msg?.text) return new Response(\"ok\");\n const chatId = String(msg.chat.id);\n\n await saveMessage(chatId, \"user\", msg.text);\n const now = new Date().toLocaleString(\"en-GB\", {\n timeZone: \"Europe/Oslo\",\n dateStyle: \"full\",\n timeStyle: \"short\",\n });\n const messages: Message[] = [\n { role: \"system\", content: SYSTEM(now) },\n ...(await history(chatId)) as Message[],\n ];\n\n const result = await runAgent({\n messages,\n tools: { ...openDataTools, ...memoryTools(chatId) },\n schemas: [...toolSchemas, ...memorySchemas],\n callModel,\n maxToolCalls: 5,\n });\n\n await saveMessage(chatId, \"assistant\", result.reply);\n await sendTelegram(chatId, result.reply || \"(no answer)\");\n return new Response(\"ok\"); // Telegram only needs a 200\n}","label",[142],"ffb7083e-a55d-47d0-b56b-0c61db2bab49","new-file-8086.ts","24f7715c-ad1b-11f1-9b90-1607ee4eb77e",6,["D",1789046752085],[152],"17bba112-50de-4ff8-9290-71c48eb80c5f","new-file-7899.ts","3319940e-ad1b-11f1-bc81-1607ee4eb77e",17,["D",1789046780479],"// Step 4 of 4 — AUTONOMY. Cron trigger, every 15 minutes (the free-plan minimum).\n// The agent wakes up without being asked, checks its own list, and messages you first.\nimport { dueReminders, markSent } from \"./memory.ts\";\nimport { sendTelegram } from \"./telegram.ts\";\n\nexport default async function () {\n const due = await dueReminders();\n for (const r of due) {\n await sendTelegram(r.chat_id, `Reminder: ${r.text}`);\n await markSent(r.id);\n }\n return `Sent ${due.length} reminder(s) at ${new Date().toISOString()}`;\n}",[158],"2cb69e86-2211-4800-8a56-15671402af9b","new-file-7256.ts","256f8b9c-ad1b-11f1-969c-1607ee4eb77e",7,["D",1789046752882],[165],"3386d44d-3059-4f6a-a85f-9301ab863117","new-file-7208.ts","2c0bc362-ad1b-11f1-a4db-1607ee4eb77e",13,["D",1789046768580],"// Step 1 of 4 — THE MODEL. No API key needed: Val.town's built-in proxy serves gpt-5.4-nano to free accounts.\n// Set OPENROUTER_API_KEY in your val's environment variables to switch to OpenRouter's free models instead.\nimport { OpenAI as ValTownOpenAI } from \"https://esm.town/v/std/openai/main.ts\";\nimport OpenAI from \"npm:openai\";\nimport type { CallModel, Message } from \"./agent.ts\";\n\nconst openrouterKey = Deno.env.get(\"OPENROUTER_API_KEY\");\nconst client = openrouterKey\n ? new OpenAI({\n baseURL: \"https://openrouter.ai/api/v1\",\n apiKey: openrouterKey,\n })\n : new ValTownOpenAI();\nexport const MODEL = openrouterKey ? \"openrouter/free\" : \"gpt-5.4-nano\";\n\nexport const callModel: CallModel = async (\n messages: Message[],\n tools: any[],\n) => {\n const completion = await client.chat.completions.create({\n model: MODEL,\n messages: messages as any,\n tools: tools.length ? tools : undefined,\n });\n return completion.choices[0].message as Message;\n};",[171],"563e1e76-0893-4d37-9cc4-0ab92e17127a","new-file-4807.ts","2881962c-ad1b-11f1-9f80-1607ee4eb77e",11,["D",1789046762583],"// Step 3 of 4 — MEMORY. Val.town gives every free account a 10 MB SQLite database.\n// Three tables: what was said, what you asked it to remember, and reminders for the cron.\nimport { sqlite } from \"https://esm.town/v/std/sqlite\";\n\nawait sqlite.batch([\n `CREATE TABLE IF NOT EXISTS messages (id INTEGER PRIMARY KEY, chat_id TEXT, role TEXT, content TEXT, created_at TEXT DEFAULT (datetime('now')))`,\n `CREATE TABLE IF NOT EXISTS notes (id INTEGER PRIMARY KEY, chat_id TEXT, note TEXT, created_at TEXT DEFAULT (datetime('now')))`,\n `CREATE TABLE IF NOT EXISTS reminders (id INTEGER PRIMARY KEY, chat_id TEXT, due_at TEXT, text TEXT, sent INTEGER DEFAULT 0)`,\n]);\n\n// libsql returns rows as arrays plus a columns list; turn them into plain objects.\nfunction rows(result: any): any[] {\n const { columns, rows } = result;\n return rows.map((\n r: any,\n ) => (Array.isArray(r)\n ? Object.fromEntries(columns.map((c: string, i: number) => [c, r[i]]))\n : r)\n );\n}\n\nexport async function history(chatId: string, limit = 20) {\n const r = await sqlite.execute({\n sql:\n `SELECT role, content FROM messages WHERE chat_id = ? ORDER BY id DESC LIMIT ?`,\n args: [chatId, limit],\n });\n return rows(r).reverse().map((m) => ({ role: m.role, content: m.content }));\n}\nexport async function saveMessage(\n chatId: string,\n role: string,\n content: string,\n) {\n await sqlite.execute({\n sql: `INSERT INTO messages (chat_id, role, content) VALUES (?, ?, ?)`,\n args: [chatId, role, content],\n });\n}\nexport async function remember(chatId: string, note: string) {\n await sqlite.execute({\n sql: `INSERT INTO notes (chat_id, note) VALUES (?, ?)`,\n args: [chatId, note],\n });\n return `Remembered: ${note}`;\n}\nexport async function recall(chatId: string) {\n const r = await sqlite.execute({\n sql:\n `SELECT note, created_at FROM notes WHERE chat_id = ? ORDER BY id DESC LIMIT 30`,\n args: [chatId],\n });\n const notes = rows(r);\n return notes.length\n ? notes.map((n) => `- ${n.note} (${n.created_at})`).join(\"\\n\")\n : \"Nothing remembered yet.\";\n}\nexport async function setReminder(\n chatId: string,\n dueAtIso: string,\n text: string,\n) {\n if (Number.isNaN(Date.parse(dueAtIso))) {\n return `I need a real date and time, got \"${dueAtIso}\".`;\n }\n await sqlite.execute({\n sql: `INSERT INTO reminders (chat_id, due_at, text) VALUES (?, ?, ?)`,\n args: [chatId, new Date(dueAtIso).toISOString(), text],\n });\n return `Reminder set for ${\n new Date(dueAtIso).toLocaleString(\"en-GB\", { timeZone: \"Europe/Oslo\" })\n }: ${text}`;\n}\nexport async function dueReminders() {\n const r = await sqlite.execute({\n sql:\n `SELECT id, chat_id, text FROM reminders WHERE sent = 0 AND due_at <= ?`,\n args: [new Date().toISOString()],\n });\n return rows(r);\n}\nexport async function markSent(id: number) {\n await sqlite.execute({\n sql: `UPDATE reminders SET sent = 1 WHERE id = ?`,\n args: [id],\n });\n}\n\n// The memory tools the model can call. They need the chat id, so we build them per request.\nexport function memoryTools(chatId: string) {\n return {\n remember: (a: any) => remember(chatId, a.note),\n recall: () => recall(chatId),\n set_reminder: (a: any) => setReminder(chatId, a.due_at, a.text),\n };\n}\nexport const memorySchemas = [\n {\n type: \"function\",\n function: {\n name: \"remember\",\n description:\n \"Store a fact about the user for later, e.g. their home stop or a preference.\",\n parameters: {\n type: \"object\",\n properties: { note: { type: \"string\" } },\n required: [\"note\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"recall\",\n description: \"List everything remembered about this user.\",\n parameters: { type: \"object\", properties: {} },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"set_reminder\",\n description:\n \"Schedule a message to the user at a future time. due_at must be an ISO 8601 datetime with timezone, e.g. 2026-09-11T07:45:00+02:00.\",\n parameters: {\n type: \"object\",\n properties: { due_at: { type: \"string\" }, text: { type: \"string\" } },\n required: [\"due_at\", \"text\"],\n },\n },\n },\n];",[178],"7d4e2d08-9b40-42c5-a889-7eb9291fb569",["D",1789046847780],2,[28],"9ac623e7-c124-4607-ada5-0045485479e1","main.ts","d1f2b354-ad1a-11f1-ac1e-1607ee4eb77e",["D",1789046612814],[189],"fffc2386-55e3-42b8-b129-998721f52ea2","agent.ts","dc9f35e8-ad1a-11f1-affb-1607ee4eb77e",5,["D",1789046735286],3,"// The agent LOOP, in 40 lines. This is the part worth reading.\n// Ask model -> if it wants a tool, run it and go round again -> stop when it answers or the budget is spent.\n// No Val.town imports here either, so it is unit-testable with a fake model.\n\nexport type Message = {\n role: \"system\" | \"user\" | \"assistant\" | \"tool\";\n content: string | null;\n tool_calls?: any[];\n tool_call_id?: string;\n name?: string;\n};\nexport type ToolFn = (args: any) => Promise;\nexport type CallModel = (messages: Message[], tools: any[]) => Promise;\n\nexport type AgentResult = {\n reply: string;\n toolCalls: { name: string; args: any; result: string }[];\n messages: Message[];\n};\n\nexport async function runAgent(opts: {\n messages: Message[]; // system + history + the new user message\n tools: Record; // name -> function\n schemas: any[]; // what the model is told about the tools\n callModel: CallModel; // the model, injected so tests can fake it\n maxToolCalls?: number; // the guardrail: autonomy without a stop is an expensive while-loop\n}): Promise {\n const budget = opts.maxToolCalls ?? 5;\n const messages = [...opts.messages];\n const log: AgentResult[\"toolCalls\"] = [];\n\n for (let round = 0; round <= budget; round++) {\n const answer = await opts.callModel(messages, opts.schemas);\n messages.push(answer);\n const wanted = answer.tool_calls ?? [];\n if (wanted.length === 0) {\n return { reply: answer.content ?? \"\", toolCalls: log, messages }; // done\n }\n\n if (round === budget) break; // budget spent: fall through to the honest exit below\n\n for (const call of wanted) {\n const name = call.function.name;\n let args: any = {};\n try {\n args = JSON.parse(call.function.arguments || \"{}\");\n } catch { /* keep {} */ }\n const fn = opts.tools[name];\n const result = fn\n ? await fn(args).catch((e) => `Tool error: ${e.message}`)\n : `Unknown tool: ${name}`;\n log.push({ name, args, result });\n messages.push({\n role: \"tool\",\n tool_call_id: call.id,\n name,\n content: result,\n });\n }\n }\n const tried = log.map((t) => t.name).join(\", \") || \"nothing\";\n return {\n reply: `I could not finish within ${budget} tool calls. I tried: ${tried}.`,\n toolCalls: log,\n messages,\n };\n}",[194],"pulls","branches","forks","envVars",{"_208":147,"_210":64},"connections",{"_208":64,"_209":64},"count","unfilledCount","emptyCount",{"_44":193,"_33":194,"_31":195,"_143":-5,"_42":43,"_41":196,"_145":25,"_39":245,"_13":198,"_34":35,"_36":199,"_24":25,"_149":-5,"_221":246,"_27":247},{"_44":188,"_33":189,"_31":190,"_143":-5,"_42":43,"_41":147,"_145":25,"_39":242,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_221":243,"_27":244},{"_44":184,"_33":28,"_31":32,"_143":-5,"_42":43,"_41":14,"_145":25,"_39":239,"_13":186,"_34":35,"_36":38,"_24":25,"_149":-5,"_221":240,"_27":241},{"_44":177,"_33":178,"_31":179,"_143":-5,"_42":43,"_41":180,"_145":25,"_39":236,"_13":147,"_34":35,"_36":182,"_24":25,"_149":-5,"_221":237,"_27":238},{"_44":170,"_33":171,"_31":172,"_143":-5,"_42":43,"_41":173,"_145":25,"_39":233,"_13":147,"_34":35,"_36":175,"_24":25,"_149":-5,"_221":234,"_27":235},{"_44":164,"_33":165,"_31":166,"_143":-5,"_42":43,"_41":167,"_145":25,"_39":230,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_221":231,"_27":232},{"_44":157,"_33":158,"_31":159,"_143":-5,"_42":43,"_41":160,"_145":25,"_39":227,"_13":147,"_34":35,"_36":162,"_24":25,"_149":-5,"_221":228,"_27":229},{"_44":151,"_33":152,"_31":153,"_143":-5,"_42":43,"_41":154,"_145":25,"_39":224,"_13":64,"_34":35,"_36":111,"_24":25,"_149":-5,"_221":225,"_27":226},{"_44":141,"_33":142,"_31":58,"_143":-5,"_42":43,"_41":144,"_145":25,"_39":220,"_13":147,"_34":35,"_36":148,"_24":25,"_149":-5,"_221":222,"_27":223},["D",1789046774524],"children",[],[142],["D",1789046752085],[],[152],["D",1789046780479],[],[158],["D",1789046752882],[],[165],["D",1789046768580],[],[171],["D",1789046762583],[],[178],["D",1789046847780],[],[28],["D",1789046612814],[],[189],["D",1789046735286],[],[194],["D",1789046612776],["D",1789488693443],{"_101":100,"_109":105},{"_101":100,"_102":103,"_104":105,"_18":56,"_106":107,"_108":25},{"_13":90,"_33":91,"_44":43,"_92":45,"_93":-5,"_94":-5,"_95":-5,"_96":-5,"_97":-5,"_98":46,"_99":100},"ctx",{"_83":-5,"_129":286,"_287":288,"_289":290,"_291":25,"_292":25,"_293":25,"_294":60,"_295":-5,"_296":-7,"_297":-7,"_298":64,"_299":25,"_300":25,"_301":-5,"_302":25,"_303":25,"_304":25,"_305":-5},"ENV",{"_257":258,"_259":260,"_261":91,"_262":263,"_264":265,"_266":267,"_268":269,"_270":271,"_272":273,"_274":275,"_276":277,"_278":279,"_280":281,"_282":283,"_284":285},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v18