ItChat

${inline(esc(para.join(\" \")))}

`;\n para = [];\n }\n };\n const flushList = () => {\n if (listType) {\n html += `<${listType}>` + listBuffer.map((li) =>\n `

${inline(esc(li))}

`\n ).join(\"\") + ``;\n listType = null;\n listBuffer = [];\n }\n };\n\n for (const raw of lines) {\n const line = raw.trimEnd();\n const heading = line.match(/^(#{1,3})\\s+(.*)$/);\n if (heading) {\n flushP();\n flushList();\n const lv = \"h\" + heading[1].length;\n html += `<${lv}>${inline(esc(heading[2]))}`;\n continue;\n }\n const ul = line.match(/^[-*]\\s+(.*)$/);\n const ol = line.match(/^\\d+\\.\\s+(.*)$/);\n if (ul || ol) {\n flushP();\n const item = ul ? ul[1]! : ol![1]!;\n const type: \"ul\" | \"ol\" = ul ? \"ul\" : \"ol\";\n if (listType === type) listBuffer.push(item);\n else {\n flushList();\n listType = type;\n listBuffer.push(item);\n }\n continue;\n }\n if (/^\\s*>\\s?/.test(line)) {\n flushP();\n flushList();\n html += `

${\n inline(esc(line.replace(/^\\s*>\\s?/, \"\")))\n }

`;\n continue;\n }\n if (line.trim() === \"\") {\n flushP();\n flushList();\n continue;\n }\n flushList();\n para.push(line);\n }\n flushP();\n flushList();\n return html;\n}\n\nfunction codeBlock(lang: string, code: string): string {\n return (\n `

` +\n `

${esc(lang)}` +\n `

` +\n `

${esc(code)}

` +\n `

`\n );\n}\n\nfunction renderInner(full: string): string {\n const regex = /```(\\w*)\\s*\\n([\\s\\S]*?)```/g;\n let out = \"\";\n let last = 0;\n let m: RegExpExecArray | null;\n while ((m = regex.exec(full))) {\n if (m.index > last) out += renderText(full.slice(last, m.index));\n out += codeBlock(m[1] || \"code\", m[2].replace(/\\n$/, \"\"));\n last = m.index + m[0].length;\n }\n if (last < full.length) out += renderText(full.slice(last));\n return out || renderText(full);\n}\n\n// --- view helpers -------------------------------------------------------------\n\nfunction assistantMsg(inner: string): string {\n return (\n `

` +\n `

${inner}

` +\n `

`\n );\n}\n\nfunction userMsg(content: string): string {\n return `

${\n inline(esc(content))\n }

`;\n}\n\nasync function main(req: Request): Promise {\n const { readable, writable } = new TransformStream();\n const writer = writable.getWriter();\n const write = (t: string) => writer.write(new TextEncoder().encode(t));\n\n (async () => {\n try {\n const form = req.method === \"POST\" ? await req.formData() : null;\n const uid = getUid(form, req);\n const key = `chat_${uid}`;\n const userText = (form?.get(\"userprompt\") as string) ?? \"\";\n\n let history: Msg[] = [];\n if (form?.get(\"clear\")) {\n await blob.delete(key);\n } else {\n history = await loadHistory(key);\n }\n\n write(`\n\n\n \n ItChat\n \n\n\n \n

`);\n\n // Render prior conversation\n for (const msg of history) {\n if (msg.role === \"user\") {\n write(userMsg(msg.content));\n } else {\n write(assistantMsg(renderInner(msg.content)));\n }\n }\n\n if (userText) {\n const nextHistory: Msg[] = [...history.slice(-MAX_HISTORY), {\n role: \"user\",\n content: userText,\n }];\n // persist the user's message so it survives even if generation fails\n await blob.setJSON(key, nextHistory);\n write(userMsg(userText));\n\n // Show a small live status while the agent runs/tests code.\n write(\n `

` +\n `

Thinking…

`,\n );\n\n const token = Deno.env.get(\"valtown\");\n if (!token) {\n throw new Error(\"The Val Town 'valtown' token is unavailable.\");\n }\n\n const modelMessages: ChatMessage[] = [\n {\n role: \"system\",\n content: SYSTEM_PROMPT +\n \" You have a run_code tool. Use it to test and debug JavaScript and TypeScript directly in this Val. \" +\n \"When a test fails, inspect the exact error, fix the code, and run it again. \" +\n \"Do not say code works unless a run succeeded. Keep the final response concise.\",\n },\n ...nextHistory,\n ];\n\n const full = await agentChat(token, modelMessages, (status) => {\n write(\n ``,\n );\n });\n\n const newHistory: Msg[] = [...nextHistory, {\n role: \"assistant\",\n content: full,\n }];\n await blob.setJSON(key, newHistory.slice(-MAX_HISTORY));\n const rendered = assistantMsg(renderInner(full));\n write(\n `` +\n ``,\n );\n }\n if (!userText && history.length === 0) {\n write(\n `

Ask me to write, run, or debug code — press Enter to send.

`,\n );\n }\n\n write(`

`);\n\n // Input area\n write(\n `

` +\n `

`,\n );\n\n write(\n ``,\n );\n\n write(``);\n } catch (e) {\n write(\n `

Something went wrong: ${\n _.escape((e && (e as any).message) || e)\n }

`,\n );\n } finally {\n writer.close();\n }\n })();\n\n return new Response(readable, { headers: { \"Content-Type\": \"text/html\" } });\n}\n\nexport default main;","deleted","label","itchat",[130],"157d0939-880a-4139-aefa-e367d5b6d105","main.tsx","01a079e9-73b0-76d8-a7b4-aa1922e8a3f8",44,["D",1788754119501],27,"import _ from \"npm:lodash@4\";\nimport { blob } from \"https://esm.town/v/std/blob/main.ts\";\n\n// OpenAI-compatible proxy. No key needed: the injected `valtown` token\n// (available in every val automatically) is used to authenticate.\nconst PROXY = \"https://std-openaiproxy.web.val.run/v1\";\n\nconst SYSTEM_PROMPT =\n \"You are VALL-E, a friendly coding chatbot. Keep chat responses concise and conversational. \" +\n \"Whenever you include code, wrap it in a fenced code block using triple backticks with a language tag, \" +\n \"like ```ts\\\\n ... \\\\n```. Only one code block per message. When finished, stop.\";\n\nconst MAX_HISTORY = 40; // cap stored messages to keep context + storage bounded\n\ntype Msg = { role: \"user\" | \"assistant\"; content: string };\n\n// --- persistence helpers -----------------------------------------------------\n\nfunction clientIp(req: Request): string {\n const cf = req.headers.get(\"cf-connecting-ip\");\n if (cf) return cf;\n const xff = req.headers.get(\"x-forwarded-for\");\n if (xff) return xff.split(\",\")[0].trim();\n return \"unknown\";\n}\n\nfunction parseCookie(h: string | null): Record {\n const o: Record = {};\n if (!h) return o;\n for (const part of h.split(\";\")) {\n const i = part.indexOf(\"=\");\n if (i > 0) {\n o[part.slice(0, i).trim()] = decodeURIComponent(part.slice(i + 1));\n }\n }\n return o;\n}\n\n// Resolve a stable per-browser id (from the form, falling back to a cookie,\n// and finally an anonymous IP key).\nfunction getUid(form: FormData | null, req: Request): string {\n const fromForm = form?.get(\"uid\") as string | null;\n if (fromForm) return fromForm;\n const cookieUid = parseCookie(req.headers.get(\"cookie\"))[\"valle_uid\"];\n if (cookieUid) return cookieUid;\n return \"anon_\" + clientIp(req);\n}\n\nasync function loadHistory(key: string): Promise {\n try {\n const v = await blob.getJSON(key);\n return Array.isArray(v) ? v : [];\n } catch {\n return [];\n }\n}\n\n// --- streaming ---------------------------------------------------------------\n\nasync function streamChat(\n token: string,\n messages: { role: string; content: string }[],\n onDelta: (text: string) => void,\n): Promise {\n const res = await fetch(`${PROXY}/chat/completions`, {\n method: \"POST\",\n headers: {\n \"content-type\": \"application/json\",\n authorization: `Bearer ${token}`,\n },\n body: JSON.stringify({\n model: \"gpt-4o\",\n messages,\n stream: true,\n max_completion_tokens: 2000,\n }),\n });\n if (!res.ok || !res.body) {\n throw new Error(\n `Model request failed: ${res.status} ${\n (await res.text().catch(() => \"\")).slice(0, 200)\n }`,\n );\n }\n\n const reader = res.body.getReader();\n const decoder = new TextDecoder();\n let buffer = \"\";\n let full = \"\";\n let sawDone = false;\n\n const processLine = (line: string) => {\n const trimmed = line.trim();\n if (!trimmed.startsWith(\"data:\")) return;\n const data = trimmed.slice(5).trim();\n if (data === \"[DONE]\") {\n sawDone = true;\n return;\n }\n try {\n const j = JSON.parse(data);\n const delta: string | undefined = j.choices?.[0]?.delta?.content;\n if (delta) {\n full += delta;\n onDelta(delta);\n }\n } catch {\n // ignore keep-alive / partial lines\n }\n };\n\n while (true) {\n const { done, value } = await reader.read();\n if (done) break;\n buffer += decoder.decode(value, { stream: true });\n let nl: number;\n while ((nl = buffer.indexOf(\"\\n\")) >= 0) {\n const line = buffer.slice(0, nl);\n buffer = buffer.slice(nl + 1);\n processLine(line);\n }\n }\n if (buffer.trim()) processLine(buffer);\n\n if (!sawDone && full) {\n throw new Error(\n \"The response was cut off mid-stream — please send your message again.\",\n );\n }\n return full;\n}\n\n// --- markdown rendering ------------------------------------------------------\n\nfunction esc(s: string): string {\n return _.escape(s);\n}\n\n// Inline formatting (bold, italic, inline code, links). Input is already escaped.\nfunction inline(s: string): string {\n s = s.replace(/`([^`]+)`/g, (_m, c: string) => `${c}`);\n s = s.replace(/\\*\\*([^*]+)\\*\\*/g, \"$1\");\n s = s.replace(/\\*([^*\\n]+)\\*/g, \"$1\");\n s = s.replace(\n /\\[([^\\]]+)\\]\$((?:https?|mailto):\\/\\/[^)\\s]+|#[A-Za-z0-9_-]+)\$/g,\n '$1',\n );\n return s;\n}\n\n// Block-level markdown for a text segment (outside code fences).\nfunction renderText(seg: string): string {\n const lines = seg.split(\"\\n\");\n let html = \"\";\n let para: string[] = [];\n let listType: \"ul\" | \"ol\" | null = null;\n let listBuffer: string[] = [];\n\n const flushP = () => {\n if (para.length) {\n html += `

${inline(esc(para.join(\" \")))}

`;\n para = [];\n }\n };\n const flushList = () => {\n if (listType) {\n html += `<${listType}>` + listBuffer.map((li) =>\n `

${inline(esc(li))}

${\n inline(esc(line.replace(/^\\s*>\\s?/, \"\")))\n }

` +\n `

${esc(lang)}` +\n `

` +\n `

${esc(code)}

` +\n `

${inner}

` +\n `

`\n );\n}\n\nfunction userMsg(content: string): string {\n return `

${\n inline(esc(content))\n }

`);\n\n // Render prior conversation\n for (const msg of history) {\n if (msg.role === \"user\") {\n write(userMsg(msg.content));\n } else {\n write(assistantMsg(renderInner(msg.content)));\n }\n }\n\n if (userText) {\n const nextHistory: Msg[] = [...history.slice(-MAX_HISTORY), {\n role: \"user\",\n content: userText,\n }];\n // persist the user's message so it survives even if generation fails\n await blob.setJSON(key, nextHistory);\n write(userMsg(userText));\n\n // NOTE: tags are intentionally left OPEN so the streamed deltas below\n // land *inside* the

. They are closed after streaming finishes.\n        write(\n          `` +\n            ``,\n        );\n\n        const token = Deno.env.get(\"valtown\");\n        const modelMessages = [\n          { role: \"system\", content: SYSTEM_PROMPT },\n          ...nextHistory,\n        ];\n\n        let full = \"\";\n        await streamChat(token, modelMessages, (d) => {\n          full += d;\n          write(_.escape(d));\n        });\n\n        // Close the tags left open above so the placeholder is a complete element.\n        write(`
`);\n\n        const newHistory: Msg[] = [...nextHistory, {\n          role: \"assistant\",\n          content: full,\n        }];\n        await blob.setJSON(key, newHistory.slice(-MAX_HISTORY));\n        const rendered = assistantMsg(renderInner(full));\n        write(\n          `` +\n            ``,\n        );\n      }\n      if (!userText && history.length === 0) {\n        write(\n          `Ask me to write some code — press Enter to send.
`,\n        );\n      }\n\n      write(`

`);\n\n // Input area\n write(\n `

` +\n `

`,\n );\n\n write(\n ``,\n );\n\n write(``);\n } catch (e) {\n write(\n `

Something went wrong: ${\n _.escape((e && (e as any).message) || e)\n }

`,\n );\n } finally {\n writer.close();\n }\n })();\n\n return new Response(readable, { headers: { \"Content-Type\": \"text/html\" } });\n}\n\nexport default main;",[147],"357a7bc2-ce65-4ccc-b895-93bd470886ab","README.md","01a079e9-73b0-76d8-a7b4-a5e9b886a96a",32,["D",1788752950189],1,"# VALL-E — coding chatbot\n\nA conversational chatbot that streams replies and returns code in copyable code blocks. No login or API key required.\n\n- Type a message; VALL-E streams its reply live.\n- Code is wrapped in fenced code blocks with a **Copy** button.\n- The conversation carries over between messages.\n\nUses the Val Town OpenAI proxy (`std-openaiproxy.web.val.run`) with the auto-injected `valtown` token, so there's nothing to configure.\n\nOpen the app: [url:main.tsx]\n",[155],"pulls","forks","envVars",{"_168":55,"_170":55},"connections",{"_168":55,"_169":55},"count","unfilledCount","emptyCount",{"_20":154,"_13":155,"_131":156,"_132":-5,"_133":21,"_134":157,"_135":51,"_22":181,"_29":159,"_138":106,"_140":160,"_142":40,"_143":-5,"_175":182,"_94":183},{"_20":146,"_13":147,"_131":148,"_132":-5,"_133":21,"_134":149,"_135":51,"_22":178,"_29":151,"_138":139,"_140":152,"_142":40,"_143":-5,"_175":179,"_94":180},{"_20":129,"_13":130,"_131":49,"_132":-5,"_133":21,"_134":30,"_135":51,"_22":174,"_29":137,"_138":139,"_140":141,"_142":40,"_143":144,"_175":176,"_94":177},["D",1788837966880],"children",[],[130],["D",1788754119501],[],[147],["D",1788752950189],[],[155],["D",1788751737705],{"_11":12,"_92":88},{"_11":12,"_85":86,"_87":88,"_46":47,"_89":90,"_91":40},{"_29":30,"_13":26,"_20":21,"_27":28,"_31":32,"_33":34,"_78":42,"_79":80,"_81":82,"_83":14,"_84":12},"ctx",{"_71":-5,"_116":221,"_222":223,"_224":225,"_226":40,"_227":40,"_228":40,"_229":51,"_230":-5,"_231":-7,"_232":-7,"_233":55,"_234":40,"_235":40,"_236":-5,"_237":40,"_238":40,"_239":40,"_240":-5},"ENV",{"_192":193,"_194":195,"_196":26,"_197":198,"_199":200,"_201":202,"_203":204,"_205":206,"_207":208,"_209":210,"_211":212,"_213":214,"_215":216,"_217":218,"_219":220},"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"]

Branches1

Branches
1