🎙️ Daily Podcast

${date}

Download MP3

\n\n`,\n { headers: { \"Content-Type\": \"text/html\" } }\n );\n}\n","label",[131],"d7b04ef2-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"import { blob } from \"https://esm.town/v/std/blob?v=12\";\n\nconst ELEVENLABS_VOICE_ID = \"21m00Tcm4TlvDq8ikWAM\"; // Rachel\n\n// ============================================================\n// DATA SOURCES — customize these for your own podcast!\n// The AI agent can add/remove/edit these via feedback.\n// ============================================================\n\n// --- Weather locations to include ---\n// Uses Open-Meteo (free, no API key needed)\n// Find coordinates at https://open-meteo.com\nconst WEATHER_LOCATIONS = [\n { name: \"New York\", lat: 40.7128, lon: -74.006 },\n // { name: \"London\", lat: 51.5074, lon: -0.1278 },\n // { name: \"Tokyo\", lat: 35.6762, lon: 139.6503 },\n];\n\n// --- Polymarket prediction markets to track ---\n// These are EVENT slugs from polymarket.com/event/{slug}\n// Add or remove as needed\nconst POLYMARKET_EVENTS: { slug: string; label: string }[] = [\n // Example: { slug: \"which-party-will-win-the-senate-in-2026\", label: \"Senate Control 2026\" },\n];\n\n// ============================================================\n// DATA FETCHERS\n// ============================================================\n\ninterface WeatherData {\n location: string;\n high: number;\n low: number;\n unit: string;\n}\n\ninterface PolymarketOutcome {\n label: string;\n question: string;\n outcomes: { name: string; probability: number }[];\n}\n\nasync function fetchWeather(): Promise {\n try {\n const weatherPromises = WEATHER_LOCATIONS.map(async (loc) => {\n const url = `https://api.open-meteo.com/v1/forecast?latitude=${loc.lat}&longitude=${loc.lon}&daily=temperature_2m_max,temperature_2m_min&timezone=auto&forecast_days=1&temperature_unit=fahrenheit`;\n const res = await fetch(url);\n const data = await res.json();\n\n const high = Math.round(data.daily.temperature_2m_max[0]);\n const low = Math.round(data.daily.temperature_2m_min[0]);\n const unit = data.daily_units.temperature_2m_max;\n\n return { location: loc.name, high, low, unit };\n });\n\n return await Promise.all(weatherPromises);\n } catch (error) {\n console.error(\"Failed to fetch weather data:\", error);\n return [];\n }\n}\n\n/**\n * Fetch prediction market data from Polymarket's Gamma API.\n *\n * Tries events endpoint first (for multi-outcome markets),\n * falls back to markets endpoint (for binary Yes/No markets).\n */\nasync function fetchPolymarketData(): Promise {\n const results: PolymarketOutcome[] = [];\n if (POLYMARKET_EVENTS.length === 0) return results;\n\n const headers = { Accept: \"application/json\", \"User-Agent\": \"ValTownDailyPodcast/1.0\" };\n\n for (const { slug, label } of POLYMARKET_EVENTS) {\n try {\n // Strategy 1: Try the events endpoint (multi-outcome markets)\n const eventRes = await fetch(\n `https://gamma-api.polymarket.com/events?slug=${slug}&closed=false`,\n { headers },\n );\n\n if (eventRes.ok) {\n const events = await eventRes.json();\n if (Array.isArray(events) && events.length > 0) {\n const event = events[0];\n const question = event.title || event.question || slug;\n const markets = event.markets || [];\n\n if (markets.length > 0) {\n const outcomes: { name: string; probability: number }[] = [];\n for (const market of markets) {\n const name = market.groupItemTitle || market.question || \"Unknown\";\n let prob = 0;\n if (market.outcomePrices) {\n try {\n const prices = JSON.parse(market.outcomePrices);\n prob = parseFloat(prices[0]) || 0;\n } catch {}\n }\n if (prob > 0.01) {\n outcomes.push({ name, probability: prob });\n }\n }\n outcomes.sort((a, b) => b.probability - a.probability);\n if (outcomes.length > 0) {\n results.push({ label, question, outcomes });\n continue;\n }\n }\n }\n }\n\n // Strategy 2: Fallback to markets endpoint (binary Yes/No)\n const marketRes = await fetch(\n `https://gamma-api.polymarket.com/markets?slug=${slug}&closed=false`,\n { headers },\n );\n\n if (marketRes.ok) {\n const markets = await marketRes.json();\n const market = Array.isArray(markets) ? markets[0] : null;\n if (market && market.outcomePrices) {\n const question = market.question || slug;\n const prices = JSON.parse(market.outcomePrices);\n const yesProb = parseFloat(prices[0]) || 0;\n if (yesProb > 0) {\n results.push({\n label,\n question,\n outcomes: [\n { name: \"Yes\", probability: yesProb },\n { name: \"No\", probability: 1 - yesProb },\n ],\n });\n }\n }\n }\n } catch (error) {\n console.error(`Failed to fetch Polymarket data for ${slug}:`, error);\n }\n }\n\n return results;\n}\n\n// ============================================================\n// PODCAST GENERATION\n// ============================================================\n\nexport async function generatePodcast(): Promise<{ url: string; text: string }> {\n const apiKey = Deno.env.get(\"ELEVEN_LABS_API_KEY\");\n if (!apiKey) throw new Error(\"ELEVEN_LABS_API_KEY not set\");\n\n const today = new Date().toLocaleDateString(\"en-US\", {\n weekday: \"long\",\n year: \"numeric\",\n month: \"long\",\n day: \"numeric\",\n timeZone: \"America/New_York\",\n });\n\n // Fetch all data sources in parallel\n const [weather, polymarket] = await Promise.all([\n fetchWeather(),\n fetchPolymarketData(),\n ]);\n\n const script = generateScript(today, weather, polymarket);\n\n // Call ElevenLabs TTS\n const res = await fetch(\n `https://api.elevenlabs.io/v1/text-to-speech/${ELEVENLABS_VOICE_ID}`,\n {\n method: \"POST\",\n headers: {\n \"xi-api-key\": apiKey,\n \"Content-Type\": \"application/json\",\n Accept: \"audio/mpeg\",\n },\n body: JSON.stringify({\n text: script,\n model_id: \"eleven_monolingual_v1\",\n voice_settings: {\n stability: 0.5,\n similarity_boost: 0.75,\n },\n }),\n },\n );\n\n if (!res.ok) {\n const err = await res.text();\n throw new Error(`ElevenLabs API error: ${res.status} ${err}`);\n }\n\n const audioBuffer = await res.arrayBuffer();\n const dateKey = new Date().toISOString().split(\"T\")[0];\n const ts = Date.now();\n const blobKey = `self-improving-podcast/episodes/${dateKey}-${ts}.mp3`;\n\n await blob.set(blobKey, new Uint8Array(audioBuffer));\n\n console.log(`Podcast generated and stored: ${blobKey} (${audioBuffer.byteLength} bytes)`);\n\n return { url: blobKey, text: script };\n}\n\n// ============================================================\n// SCRIPT BUILDER\n// ============================================================\n\nfunction generateScript(\n today: string,\n weather?: WeatherData[],\n polymarket?: PolymarketOutcome[],\n): string {\n const parts: string[] = [\n `Good morning. It's ${today}.`,\n \"Here's your daily brief.\",\n ];\n\n // Weather\n if (weather && weather.length > 0) {\n parts.push(\"Let's start with the weather.\");\n for (const loc of weather) {\n parts.push(\n `In ${loc.location}, expect a high of ${loc.high} and a low of ${loc.low} degrees Fahrenheit.`,\n );\n }\n }\n\n // Prediction markets\n if (polymarket && polymarket.length > 0) {\n parts.push(\"Now, prediction markets.\");\n for (const market of polymarket) {\n if (\n market.outcomes.length > 2\n || (market.outcomes.length === 2 && market.outcomes[0].name !== \"Yes\")\n ) {\n const top = market.outcomes.slice(0, 3);\n const outcomeText = top\n .map((o) => `${o.name} at ${Math.round(o.probability * 100)}%`)\n .join(\", \");\n parts.push(`${market.label}: Polymarket has ${outcomeText}.`);\n } else {\n const pct = Math.round(market.outcomes[0].probability * 100);\n parts.push(`${market.label}: Polymarket says ${pct}% yes.`);\n }\n }\n }\n\n parts.push(\"Have a great day.\");\n\n return parts.join(\" \");\n}\n",[126],"d7ae86ee-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"import { blob } from \"https://esm.town/v/std/blob?v=12\";\nimport { generatePodcast } from \"./generate-podcast.ts\";\nimport { runAgent } from \"./agent.ts\";\n\n/** Process any unprocessed podcast feedback through the agent */\nasync function processPendingFeedback() {\n try {\n const allFeedback = await blob.list(\"self-improving-podcast/feedback/\");\n const unprocessed: { key: string; data: any }[] = [];\n\n for (const f of allFeedback) {\n const data = await blob.getJSON(f.key) as any;\n if (data && !data.processed) {\n unprocessed.push({ key: f.key, data });\n }\n }\n\n if (unprocessed.length === 0) {\n console.log(\"No pending feedback to process.\");\n return;\n }\n\n console.log(`Processing ${unprocessed.length} pending feedback items...`);\n\n // Combine all feedback into one agent call\n const feedbackText = unprocessed\n .map((f, i) => `Feedback ${i + 1} (${f.data.episode || \"general\"}):\\n${f.data.feedback}`)\n .join(\"\\n\\n\");\n\n const result = await runAgent(\n `Process the following user feedback for the podcast:\\n\\n${feedbackText}`,\n \"podcast-feedback-batch\",\n );\n\n console.log(\"Agent processed feedback:\", result.appliedEdits.length, \"edits applied\");\n\n // Mark all as processed\n for (const f of unprocessed) {\n await blob.setJSON(f.key, { ...f.data, processed: true, processedAt: new Date().toISOString() });\n }\n } catch (err) {\n console.error(\"Failed to process feedback:\", err.message);\n // Don't block episode generation if feedback processing fails\n }\n}\n\nexport default async function () {\n const today = new Date().toLocaleDateString(\"en-US\", {\n weekday: \"long\",\n year: \"numeric\",\n month: \"long\",\n day: \"numeric\",\n timeZone: \"America/New_York\",\n });\n\n // Step 1: Process any pending feedback through the agent\n // This may modify generate-podcast.ts before we call it\n await processPendingFeedback();\n\n // Step 2: Generate podcast\n try {\n const podcast = await generatePodcast();\n console.log(`✅ Podcast generated for ${today}: ${podcast.url}`);\n console.log(`Script: ${podcast.text.slice(0, 200)}...`);\n } catch (err) {\n console.error(`❌ Podcast generation failed for ${today}:`, err.message);\n }\n}\n",[119],"d7ae40e4-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"import { blob } from \"https://esm.town/v/std/blob?v=12\";\n\n// ============================================================\n// PODCAST METADATA — customize these for your podcast!\n// ============================================================\nconst PODCAST_TITLE = \"My Daily Brief\";\nconst PODCAST_DESCRIPTION = \"A personal daily audio briefing, generated by a self-improving Val Town system.\";\nconst PODCAST_AUTHOR = \"Daily Brief Bot\";\n\nexport default async function(req: Request): Promise {\n const url = new URL(req.url);\n const baseUrl = url.origin;\n\n // Serve audio: /episode/2026-03-16-1234567890.mp3 or /episode/2026-03-16.mp3 (legacy)\n const episodeMatch = url.pathname.match(/^\\/episode\\/(\\d{4}-\\d{2}-\\d{2}(?:-\\d+)?)\\.mp3$/);\n if (episodeMatch) {\n const key = episodeMatch[1];\n const blobKey = `self-improving-podcast/episodes/${key}.mp3`;\n try {\n const audioResponse = await blob.get(blobKey);\n if (!audioResponse) return new Response(\"Not found\", { status: 404 });\n const audioBytes = new Uint8Array(await audioResponse.arrayBuffer());\n const total = audioBytes.byteLength;\n\n // Handle Range requests (Apple Podcasts needs this for streaming)\n const rangeHeader = req.headers.get(\"Range\");\n if (rangeHeader) {\n const match = rangeHeader.match(/bytes=(\\d+)-(\\d*)/);\n if (match) {\n const start = parseInt(match[1]);\n const end = match[2] ? Math.min(parseInt(match[2]), total - 1) : total - 1;\n const chunk = audioBytes.slice(start, end + 1);\n return new Response(chunk, {\n status: 206,\n headers: {\n \"Content-Type\": \"audio/mpeg\",\n \"Content-Length\": String(chunk.byteLength),\n \"Content-Range\": `bytes ${start}-${end}/${total}`,\n \"Accept-Ranges\": \"bytes\",\n \"Cache-Control\": \"public, max-age=86400\",\n },\n });\n }\n }\n\n return new Response(audioBytes, {\n headers: {\n \"Content-Type\": \"audio/mpeg\",\n \"Content-Length\": String(total),\n \"Accept-Ranges\": \"bytes\",\n \"Cache-Control\": \"public, max-age=86400\",\n },\n });\n } catch {\n return new Response(\"Not found\", { status: 404 });\n }\n }\n\n // Serve cover art: /cover.svg\n if (url.pathname === \"/cover.svg\") {\n const svg = ``;\n return new Response(svg, { headers: { \"Content-Type\": \"image/svg+xml\", \"Cache-Control\": \"public, max-age=86400\" } });\n }\n\n // Feedback form: GET /feedback or /feedback?ep=2026-03-17\n if (url.pathname === \"/feedback\") {\n const ep = url.searchParams.get(\"ep\") || \"general\";\n\n if (req.method === \"POST\") {\n const form = await req.formData();\n const feedback = form.get(\"feedback\")?.toString() || \"\";\n\n if (feedback.trim()) {\n // Save feedback — this is fast, just a blob write\n const key = `self-improving-podcast/feedback/${Date.now()}.json`;\n await blob.setJSON(key, {\n episode: ep,\n feedback: feedback.trim(),\n submittedAt: new Date().toISOString(),\n processed: false,\n });\n }\n\n // Return immediately — agent runs later during daily cron\n return new Response(feedbackPage(baseUrl, ep, true), {\n headers: { \"Content-Type\": \"text/html\" },\n });\n }\n\n return new Response(feedbackPage(baseUrl, ep, false), {\n headers: { \"Content-Type\": \"text/html\" },\n });\n }\n\n // Default: serve RSS feed\n const feedUrl = `${baseUrl}/`;\n\n const allBlobs = await blob.list(\"self-improving-podcast/episodes/\");\n const episodes = allBlobs\n .filter((b: any) => b.key.endsWith(\".mp3\"))\n .map((b: any) => {\n const dateMatch = b.key.match(/(\\d{4}-\\d{2}-\\d{2})(?:-(\\d+))?\\.mp3$/);\n if (!dateMatch) return null;\n const date = dateMatch[1];\n const ts = dateMatch[2] || \"0\";\n const filename = b.key.split(\"/\").pop()!.replace(\".mp3\", \"\");\n return { date, ts, filename, size: b.size };\n })\n .filter(Boolean)\n .sort((a: any, b: any) => {\n const d = b.date.localeCompare(a.date);\n if (d !== 0) return d;\n return b.ts.localeCompare(a.ts);\n });\n\n const items = episodes\n .map((ep: any) => {\n const audioUrl = `${baseUrl}/episode/${ep.filename}.mp3`;\n const feedbackUrl = `${baseUrl}/feedback?ep=${ep.date}`;\n const pubDate = ep.ts !== \"0\"\n ? new Date(parseInt(ep.ts)).toUTCString()\n : new Date(ep.date + \"T08:00:00Z\").toUTCString();\n const guid = `daily-brief-${ep.filename}`;\n return ` \n Daily Brief ${ep.date}\n Your daily audio briefing for ${ep.date}. Leave feedback for the next episode: ${feedbackUrl}\n ${pubDate}\n \n ${guid}\n 60\n `;\n })\n .join(\"\\n\");\n\n const rss = `\n\n \n ${esc(PODCAST_TITLE)}\n ${esc(PODCAST_DESCRIPTION)}. Leave feedback: ${baseUrl}/feedback\n en\n ${esc(PODCAST_AUTHOR)}\n \n false\n \n ${esc(feedUrl)}\n${items}\n \n`;\n\n return new Response(rss, {\n headers: {\n \"Content-Type\": \"application/rss+xml; charset=utf-8\",\n \"Cache-Control\": \"public, max-age=300\",\n },\n });\n}\n\nfunction feedbackPage(baseUrl: string, ep: string, submitted: boolean): string {\n const resultHtml = submitted\n ? `

Got it! Your feedback has been saved and the AI agent will process it before the next episode.

`\n : \"\";\n\n return `\n\n\n Podcast Feedback — ${ep}\n \n \n\n\n

🎙️ Podcast Feedback

${ep === \"general\" ? \"General feedback\" : `Episode: ${ep}`}

\n ${resultHtml}\n \n

Your feedback is saved instantly and processed by an AI agent before the next episode generates.

\n\n`;\n}\n\nfunction esc(s: string): string {\n return s.replace(/&/g, \"&\").replace(//g, \">\").replace(/\"/g, \""\");\n}\n",[113],"d7ae36e4-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"# Feature Ideas\n\nThings to build eventually. Roughly ordered by excitement.\n\n---\n\n## More data sources\n- Calendar summary for the day\n- News headlines (top 3)\n- Stock portfolio performance\n- Habit tracker check-ins\n\n## Smarter script generation\n- Use an LLM to write the script instead of string templates\n- More natural transitions between sections\n- Vary the opening/closing each day\n\n## Multi-voice conversations\n- Use two ElevenLabs voices for a host + co-host format\n- One voice reads data, the other comments\n\n## Episode show notes\n- Auto-generate markdown show notes alongside each episode\n- Include links to data sources mentioned\n\n## Self-evaluation\n- After making changes, the agent could run the val and check for errors\n- Auto-revert if something breaks\n",[107],"d7ae2c76-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"// Redirects to a placeholder cover art image (1400x1400, Apple Podcasts minimum)\nexport default async function(req: Request): Promise {\n // Unsplash image cropped to 1400x1400 — sky and clouds\n const imageUrl = \"https://images.unsplash.com/photo-1506905925346-21bda4d32df4?w=1400&h=1400&fit=crop&crop=center\";\n\n return Response.redirect(imageUrl, 302);\n}\n",[100],"d7ae1f6a-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"# Changelog\n\nAll changes made by the AI agent via feedback are logged here.\n\n---\n\n## 2026-03-25 — Initial version\n- Created self-improving-podcast val\n- generate-daily.ts: interval that processes feedback + generates podcast\n- generate-podcast.ts: weather (Open-Meteo) + optional Polymarket → ElevenLabs TTS\n- feed.ts: RSS feed, episode serving, feedback form\n- agent.ts: self-improving agent reads own source, calls Claude, applies edits\n- podcast-player.ts: simple web audio player\n- cover-art.ts: placeholder cover art redirect\n",[94],"d7ad57ce-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"/**\n * Shared Townie agent: reads val source, sends to Claude, applies edits.\n * Used by podcast feedback to self-improve.\n */\nimport Anthropic from \"npm:@anthropic-ai/sdk@0.39.0\";\nimport { blob } from \"https://esm.town/v/std/blob?v=12\";\n\n// IMPORTANT: V2 API needs the UUID, not handle/valName\nconst VAL_UUID = \"9af62dd2-288f-11f1-b4ab-42dde27851f2\";\n\ninterface FileInfo {\n path: string;\n content: string;\n}\n\ninterface Edit {\n path: string;\n action: string;\n content: string;\n}\n\nexport interface AgentResult {\n aiText: string;\n appliedEdits: string[];\n}\n\nexport async function runAgent(userMessage: string, context?: string): Promise {\n const token = Deno.env.get(\"TOWNIE_TOKEN\");\n if (!token) throw new Error(\"TOWNIE_TOKEN not set\");\n\n // Read val's own files\n const files = await fetchValFiles(token);\n console.log(`Agent read ${files.length} files`);\n\n // Call Claude\n const anthropic = new Anthropic();\n const systemPrompt = buildSystemPrompt(files, context);\n\n const response = await anthropic.messages.create({\n model: \"claude-sonnet-4-20250514\",\n max_tokens: 4096,\n system: systemPrompt,\n messages: [\n { role: \"user\", content: userMessage },\n ],\n });\n\n const aiText = response.content\n .filter((b) => b.type === \"text\")\n .map((b) => b.text)\n .join(\"\\n\");\n\n console.log(\"Agent response:\", aiText.slice(0, 500));\n\n // Parse and apply edits\n const edits = parseEdits(aiText);\n const appliedEdits: string[] = [];\n\n for (const edit of edits) {\n try {\n await applyEdit(token, edit);\n appliedEdits.push(`✅ ${edit.action} ${edit.path}`);\n } catch (err) {\n appliedEdits.push(`❌ ${edit.action} ${edit.path}: ${err.message}`);\n }\n }\n\n // Update changelog\n if (appliedEdits.length > 0) {\n await appendChangelog(token, userMessage.slice(0, 100), appliedEdits);\n }\n\n // Archive\n await blob.setJSON(\n `self-improving-podcast/agent-runs/${new Date().toISOString()}.json`,\n {\n source: context || \"unknown\",\n message: userMessage,\n reply: aiText,\n edits: appliedEdits,\n timestamp: new Date().toISOString(),\n }\n );\n\n return { aiText, appliedEdits };\n}\n\nasync function fetchValFiles(token: string): Promise {\n // List all files (path= with empty string lists root)\n const res = await fetch(\n `https://api.val.town/v2/vals/${VAL_UUID}/files?path=`,\n { headers: { Authorization: `Bearer ${token}` } }\n );\n if (!res.ok) throw new Error(`Failed to list files: ${res.status} ${await res.text()}`);\n const data = await res.json();\n const files: FileInfo[] = [];\n\n for (const file of data.data || []) {\n if (file.type === \"directory\") continue;\n if (!file.path.endsWith(\".ts\") && !file.path.endsWith(\".md\")) continue;\n if (file.path.includes(\"debug\") || file.path.includes(\"check-episodes\")) continue;\n\n // Read content via esm.town module URL\n const moduleUrl = file.links?.module;\n if (!moduleUrl) continue;\n\n try {\n const contentRes = await fetch(moduleUrl);\n if (!contentRes.ok) continue;\n const content = await contentRes.text();\n files.push({ path: file.path, content });\n } catch {\n console.error(`Failed to read ${file.path}`);\n }\n }\n\n return files;\n}\n\nfunction buildSystemPrompt(files: FileInfo[], context?: string): string {\n const fileList = files\n .map((f) => `--- ${f.path} ---\\n${f.content}`)\n .join(\"\\n\\n\");\n\n return `You are an AI assistant that helps improve a Val Town project called \"self-improving-podcast\".\nThis val generates a daily podcast and serves it via RSS feed. It can be improved by user feedback.\n\n${context ? `Context: ${context}\\n\\n` : \"\"}\n\nHere are the current files in the val:\n\n${fileList}\n\nRULES:\n- If the user asks for a code change, output edit blocks (see format below).\n- If they're just chatting or asking a question, just respond naturally.\n- Keep changes minimal and focused.\n- Always explain what you're doing.\n- Never delete agent.ts or break the self-editing loop.\n- For podcast content changes (new data sources, topics), edit generate-podcast.ts.\n- For scheduling changes, edit generate-daily.ts.\n- For feed/player changes, edit feed.ts or podcast-player.ts.\n- Use free APIs that don't require API keys when possible (e.g. open-meteo.com for weather, not openweathermap).\n- The val runs on Deno, so use npm: prefix for npm packages.\n\nEDIT FORMAT:\nTo create or replace a file, use:\n<<>>\nfile content here\n<<>>\n\nTo create a new file:\n<<>>\nfile content here\n<<>>\n\nOnly output edit blocks if you're actually making changes.`;\n}\n\nfunction parseEdits(text: string): Edit[] {\n const edits: Edit[] = [];\n const regex =\n /<<>>\\n([\\s\\S]*?)<<>>/g;\n let match;\n while ((match = regex.exec(text)) !== null) {\n edits.push({ path: match[1], action: match[2], content: match[3].trim() });\n }\n return edits;\n}\n\nasync function applyEdit(token: string, edit: Edit) {\n const url = `https://api.val.town/v2/vals/${VAL_UUID}/files`;\n\n let res: Response;\n if (edit.action === \"create\") {\n res = await fetch(url, {\n method: \"POST\",\n headers: {\n Authorization: `Bearer ${token}`,\n \"Content-Type\": \"application/json\",\n },\n body: JSON.stringify({\n path: edit.path,\n content: edit.content,\n type: \"script\",\n }),\n });\n } else {\n res = await fetch(`${url}?path=${encodeURIComponent(edit.path)}`, {\n method: \"PUT\",\n headers: {\n Authorization: `Bearer ${token}`,\n \"Content-Type\": \"application/json\",\n },\n body: JSON.stringify({ content: edit.content }),\n });\n }\n\n if (!res.ok) {\n const body = await res.text();\n throw new Error(`API ${res.status}: ${body.slice(0, 200)}`);\n }\n\n console.log(`Edit applied: ${edit.action} ${edit.path}`);\n}\n\nasync function appendChangelog(token: string, request: string, edits: string[]) {\n const date = new Date().toISOString().split(\"T\")[0];\n const entry = `\\n## ${date} — via agent\\n- Request: \"${request}\"\\n${edits.map((e) => `- ${e}`).join(\"\\n\")}\\n`;\n\n try {\n const res = await fetch(\n `https://api.val.town/v2/vals/${VAL_UUID}/files?path=changelog.md`,\n { headers: { Authorization: `Bearer ${token}` } }\n );\n // Read existing content via module URL\n const data = await res.json();\n const moduleUrl = data.data?.[0]?.links?.module;\n let current = \"# Changelog\\n\";\n if (moduleUrl) {\n const contentRes = await fetch(moduleUrl);\n if (contentRes.ok) current = await contentRes.text();\n }\n\n const putRes = await fetch(\n `https://api.val.town/v2/vals/${VAL_UUID}/files?path=changelog.md`,\n {\n method: \"PUT\",\n headers: {\n Authorization: `Bearer ${token}`,\n \"Content-Type\": \"application/json\",\n },\n body: JSON.stringify({ content: current + entry }),\n }\n );\n\n if (!putRes.ok) {\n console.error(\"Failed to update changelog:\", putRes.status);\n }\n } catch (err) {\n console.error(\"Changelog update error:\", err.message);\n }\n}\n",[87],"d7acd380-29a6-11f1-9f45-42dde27851f2",["D",1774593198355],"# 🎙️ Self-Improving Podcast\n\nA daily podcast that generates itself — and improves itself based on your feedback.\n\nFork this val, add your API keys, and you have a personal daily audio briefing that gets better over time.\n\n## How it works\n\n1. **`generate-daily.ts`** runs on a schedule (interval)\n2. It checks for any unprocessed feedback and feeds it to the **AI agent**\n3. The agent can edit the val's own source code (e.g. add new data sources, change the script format)\n4. Then **`generate-podcast.ts`** fetches live data and generates a TTS episode via ElevenLabs\n5. The episode is stored as a blob and served via **`feed.ts`** as an RSS feed\n\nSubscribe in your podcast app. Leave feedback via the web form. The AI agent processes it before the next episode.\n\n## Files\n\n| File | Type | What it does |\n|------|------|-------------|\n| `generate-daily.ts` | interval | Cron trigger — processes feedback, then generates today's episode |\n| `generate-podcast.ts` | script | Fetches data sources, builds script, calls ElevenLabs TTS |\n| `feed.ts` | http | RSS feed, episode audio serving, feedback form (`/feedback`) |\n| `podcast-player.ts` | http | Simple web audio player |\n| `agent.ts` | script | Self-improving AI agent — reads own source, calls Claude, applies code edits |\n| `cover-art.ts` | http | Podcast cover art |\n| `changelog.md` | file | Log of all agent-applied changes |\n| `feature-ideas.md` | file | Future ideas |\n\n## Built-in data sources\n\nOut of the box, the podcast script includes:\n\n- **Weather** — via [Open-Meteo](https://open-meteo.com) (free, no API key). Default: New York. Edit `WEATHER_LOCATIONS` in `generate-podcast.ts` to change.\n- **Prediction markets** — via [Polymarket Gamma API](https://gamma-api.polymarket.com) (free, no API key). Empty by default — uncomment or add event slugs in `POLYMARKET_EVENTS`.\n\nWant more? Just submit feedback like _\"Add top Hacker News stories\"_ or _\"Include Bitcoin price\"_ and the agent will wire it up.\n\n## Setup\n\n1. **Fork this val**\n2. **Add environment variables:**\n\n| Key | Description |\n|-----|-------------|\n| `ELEVEN_LABS_API_KEY` | [ElevenLabs](https://elevenlabs.io) text-to-speech API key |\n| `ANTHROPIC_API_KEY` | [Anthropic](https://anthropic.com) API key for Claude (used by agent) |\n| `TOWNIE_TOKEN` | Val Town API token with write access (used by agent to edit its own files) |\n\n3. **Update `agent.ts`** — change `VAL_UUID` to your forked val's UUID\n4. **Set the interval schedule** on `generate-daily.ts` (e.g. daily at 7am)\n5. **Subscribe** to the RSS feed URL in your podcast app (see below)\n\n## Listening\n\nYour feed URL is the endpoint for `feed.ts` — copy it from the Val Town dashboard.\n\n### Apple Podcasts (iPhone / iPad)\n\n1. Open the **Podcasts** app\n2. Tap **Library** at the bottom\n3. Tap the **⋯** (three dots) in the top-right\n4. Tap **Follow a Show by URL**\n5. Paste your feed URL → tap **Follow**\n\n### Apple Podcasts (Mac)\n\n1. Open **Podcasts** (or iTunes on older macOS)\n2. Click **File** → **Add a Show by URL…**\n3. Paste your feed URL → click **Follow**\n\n### Overcast (iOS)\n\n1. Tap the magnifying glass icon\n2. Tap **Add URL**\n3. Paste your feed URL\n\n### Pocket Casts\n\n1. Go to the **Discover** tab\n2. Paste your feed URL into the search bar\n3. Tap **Subscribe**\n\n### Podcast Addict (Android)\n\n1. Tap **+** → **RSS feed**\n2. Paste your feed URL\n\n### Spotify\n\n⚠️ Spotify does **not** support subscribing to private/custom RSS feeds. You'd need to submit your podcast publicly through [Spotify for Creators](https://creators.spotify.com), which isn't really the point here. Use one of the apps above instead.\n\n## Feedback loop\n\n1. Visit `/feedback` on the feed endpoint (linked in each episode description)\n2. Submit text feedback — it's saved as a blob instantly\n3. Next time `generate-daily.ts` runs, the agent reads all unprocessed feedback\n4. The agent calls Claude with the full val source + your feedback\n5. Claude outputs edit blocks → agent applies them via the Val Town API\n6. Changes are logged in `changelog.md`\n7. The (potentially updated) `generate-podcast.ts` runs to create that day's episode\n\n## Customization\n\nEdit the constants at the top of each file:\n\n- **`generate-podcast.ts`** — `WEATHER_LOCATIONS`, `POLYMARKET_EVENTS`, `ELEVENLABS_VOICE_ID`\n- **`feed.ts`** — `PODCAST_TITLE`, `PODCAST_DESCRIPTION`, `PODCAST_AUTHOR`\n\nOr just submit feedback and let the agent do it for you!\n\n## Blob storage layout\n\n```\nself-improving-podcast/\n episodes/ — MP3 files (2026-03-25-1711234567890.mp3)\n feedback/ — Feedback JSON blobs (timestamp.json)\n agent-runs/ — Agent execution logs\n```\n",[76],"pulls","branches","forks","envVars",{"_127":20,"_270":20},"connections",{"_127":20,"_269":20},"unfilledCount","emptyCount",{"_27":258,"_13":76,"_74":75,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":305,"_33":20,"_77":48,"_226":260,"_78":24,"_228":-5,"_281":306,"_180":307},{"_27":254,"_13":87,"_74":86,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":302,"_33":20,"_77":88,"_226":256,"_78":24,"_228":-5,"_281":303,"_180":304},{"_27":250,"_13":94,"_74":93,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":299,"_33":20,"_77":48,"_226":252,"_78":24,"_228":-5,"_281":300,"_180":301},{"_27":246,"_13":100,"_74":99,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":296,"_33":20,"_77":101,"_226":248,"_78":24,"_228":-5,"_281":297,"_180":298},{"_27":242,"_13":107,"_74":106,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":293,"_33":20,"_77":48,"_226":244,"_78":24,"_228":-5,"_281":294,"_180":295},{"_27":238,"_13":113,"_74":112,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":290,"_33":20,"_77":101,"_226":240,"_78":24,"_228":-5,"_281":291,"_180":292},{"_27":234,"_13":119,"_74":118,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":287,"_33":20,"_77":120,"_226":236,"_78":24,"_228":-5,"_281":288,"_180":289},{"_27":230,"_13":126,"_74":125,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":284,"_33":20,"_77":88,"_226":232,"_78":24,"_228":-5,"_281":285,"_180":286},{"_27":222,"_13":131,"_74":130,"_37":-5,"_29":30,"_223":34,"_224":200,"_31":280,"_33":20,"_77":101,"_226":227,"_78":24,"_228":-5,"_281":282,"_180":283},["D",1774593198355],"children",[],[131],["D",1774593198355],[],[126],["D",1774593198355],[],[119],["D",1774593198355],[],[113],["D",1774593198355],[],[107],["D",1774593198355],[],[100],["D",1774593198355],[],[94],["D",1774593198355],[],[87],["D",1774593198355],[],[76],["D",1774593198355],{"_11":12,"_179":81},{"_11":12,"_175":-5,"_80":81,"_25":79,"_176":177,"_178":24},{"_33":34,"_13":16,"_27":30,"_165":132,"_166":71,"_167":72,"_168":136,"_169":170,"_171":172,"_173":14,"_174":12},"ctx",{"_46":-5,"_202":345,"_346":347,"_348":349,"_350":24,"_351":24,"_352":24,"_353":200,"_354":-5,"_355":-7,"_356":-7,"_357":20,"_358":24,"_359":24,"_360":-5,"_361":24,"_362":24,"_363":24,"_364":-5},"ENV",{"_316":317,"_318":319,"_320":16,"_321":322,"_323":324,"_325":326,"_327":328,"_329":330,"_331":332,"_333":334,"_335":336,"_337":338,"_339":340,"_341":342,"_343":344},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","71c1b20804b3232a7e513be260cfffc1932a9e97","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