[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_200":201,"_202":203},"routes/_app.x.$handle.$name",{"_93":92,"_31":43,"_102":83,"_26":103,"_13":14,"_104":-5,"_11":105,"_106":107,"_108":109,"_110":111,"_112":113,"_20":-5,"_114":-5,"_115":-5,"_116":24,"_117":118,"_119":120,"_121":-5,"_122":-5,"_123":124},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_41":42,"_31":43,"_36":44,"_45":-5,"_46":24,"_47":48,"_49":48,"_50":51,"_52":-5,"_18":53,"_54":-5,"_55":24,"_56":-5,"_57":-5,"_58":59,"_60":61,"_62":-5,"_63":-5,"_64":-5,"_65":-5,"_66":67,"_68":67,"_69":70,"_71":70,"_72":70,"_73":67,"_74":67,"_75":76,"_77":78,"_79":80,"_81":24},"version",5,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":-7,"_23":24,"_25":24,"_26":27,"_28":27},"file",{"_29":30,"_31":27,"_32":33,"_34":35,"_36":37,"_38":14,"_39":40,"_23":24},"previous","deleted",false,"deletedHere","path","articles-demo.ts","latestPath","fileId","50fc6998-8691-11f1-a028-1607ee4eb77e","name","fileType","script","content","/**\n * Runnable harness for articles-effect.ts: mock Database/Notifications, a\n * standalone Effect v4 + OTel runtime, and an HTTP-friendly entry point.\n *\n * Deliberately a separate Effect major version from main.ts (which runs\n * effect@3.22.0 for the writing-assistant demo). Effect v3 and v4 values\n * cannot cross a pipe/provide boundary, so this file owns its own\n * ManagedRuntime end to end; only the resulting Promise crosses into main.ts.\n */\nimport { Duration, Effect, Layer, LogLevel, ManagedRuntime, References } from \"https://esm.sh/effect@4.0.0-beta.101\"\nimport * as NodeSdk from \"https://esm.sh/@effect/opentelemetry@4.0.0-beta.101/NodeSdk\"\nimport { SimpleSpanProcessor } from \"https://esm.sh/@opentelemetry/sdk-trace-base@2\"\nimport { Articles, Database, handleRequest, Notifications } from \"./articles-effect.ts\"\nimport { TreeSpanExporter } from \"./trace-tree.ts\"\n\nconst NodeSdkLive = NodeSdk.layer(() => ({\n resource: { serviceName: \"articles-effect-demo\" },\n spanProcessor: new SimpleSpanProcessor(new TreeSpanExporter()),\n}))\n\n// In-memory stand-in for drizzle: an incrementing id, a small artificial delay.\nlet nextId = 1\nconst DatabaseDemo = Layer.succeed(Database, {\n insertArticle: (values) =>\n Effect.succeed({ id: nextId++, slug: values.slug, title: values.title, authorId: values.authorId }).pipe(\n Effect.delay(Duration.millis(120)),\n ),\n})\n\n// `fail` lets the presenter trigger the notification-failure path live,\n// on demand, instead of leaving it to chance.\nconst notificationsLayer = (fail: boolean) =>\n Layer.succeed(Notifications, {\n articleCreated: () =>\n (fail ? Effect.fail(new Error(\"notification service unreachable\")) : Effect.void).pipe(\n Effect.delay(Duration.millis(400)),\n ),\n })\n\n// The ✗ on a failed span already shows the failure in the trace tree; the\n// console WARN from Effect.logWarning would just be noise on top of it.\nconst runtime = ManagedRuntime.make(\n Layer.merge(NodeSdkLive, Layer.succeed(References.MinimumLogLevel, LogLevel.None)),\n)\n\nexport interface ArticleDemoInput {\n readonly title: string\n readonly body?: string\n readonly authorId?: number\n readonly failNotification?: boolean\n readonly tenant?: { id: string; region: string }\n}\n\nexport function runArticleDemo(input: ArticleDemoInput) {\n const tenant = input.tenant ?? { id: \"demo-tenant\", region: \"eu-west\" }\n const ArticlesLive = Articles.layerNoDeps.pipe(\n Layer.provide(Layer.merge(DatabaseDemo, notificationsLayer(!!input.failNotification))),\n )\n\n return runtime.runPromise(\n handleRequest(tenant, input.authorId ?? 1, { title: input.title, body: input.body ?? \"\" }).pipe(\n Effect.provide(ArticlesLive),\n Effect.withSpan(\"http /demo/article\"),\n ),\n )\n}\n","createdAt",["D",1784809411130],"branchVersion","branchId","1b4f9e5b-2b97-4454-af55-2b8d4937cd3c","id","9d37c6fa-1f6b-4e6c-b37d-362632a15818","effect-utrecht-meetup",["D",1784809347061],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","Demo code from the Effect-TS Meetup Utrecht talk","imageUrl","0b86c720-ddc6-11ef-94bf-d1c3ffb48162","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","7018ec7d-7a8e-4e8f-93d1-ab6a7334ca50","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_93":92,"_101":97},"org",{"_93":92,"_94":95,"_96":97,"_18":53,"_98":99,"_100":24},"branch",{"_13":82,"_31":83,"_41":40,"_84":42,"_85":-5,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":43,"_91":92},"isOwned",9,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","cricks_unmixed4u","handle","displayName","ozan","avatar","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzJzSEJuSGdYT2dQSmprejVnU0VqalpDaWhKUSJ9","tier","pro","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_41":42,"_31":43,"_36":196,"_45":-5,"_46":24,"_47":48,"_49":48,"_50":51,"_52":-5,"_18":53,"_54":-5,"_55":24,"_56":-5,"_57":-5,"_58":59,"_60":61,"_62":-5,"_63":-5,"_64":-5,"_65":-5,"_66":67,"_68":67,"_69":70,"_71":70,"_72":70,"_73":67,"_74":67,"_75":197,"_77":198,"_79":199,"_81":24},"tree",[175,176,177,178,179],"files",[175,176,177,178,179],"counts",{"_165":59,"_166":143,"_167":59,"_168":169,"_170":171},"contents",[125,126,127,128,129],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles",true,"grants","orgs","customDomains",{},{"_41":162,"_31":27,"_29":30,"_133":-5,"_39":40,"_38":14,"_135":120,"_36":163,"_13":59,"_32":33,"_34":35,"_23":24,"_138":-5,"_26":164},{"_41":155,"_31":156,"_29":157,"_133":-5,"_39":40,"_38":158,"_135":120,"_36":159,"_13":59,"_32":33,"_34":160,"_23":24,"_138":-5,"_26":161},{"_41":148,"_31":149,"_29":150,"_133":-5,"_39":40,"_38":151,"_135":24,"_36":152,"_13":59,"_32":145,"_34":153,"_23":24,"_138":-5,"_26":154},{"_41":140,"_31":141,"_29":142,"_133":-5,"_39":40,"_38":143,"_135":24,"_36":144,"_13":59,"_32":145,"_34":146,"_23":24,"_138":-5,"_26":147},{"_41":130,"_31":131,"_29":132,"_133":-5,"_39":40,"_38":134,"_135":120,"_36":136,"_13":59,"_32":33,"_34":137,"_23":24,"_138":-5,"_26":139},"9bf274e7-a055-4524-9382-12812fd90493","trace-tree.ts","49b5bc84-8691-11f1-b267-1607ee4eb77e","parentId",4,"latest",["D",1784809399012],"/**\n * A tiny SpanExporter that, instead of dumping raw OTel JSON, records spans\n * in memory so we can print them as an indented ASCII tree: parent/child\n * nesting and sibling concurrency made visible at a glance, for live demos.\n */\n\nexport interface RecordedSpan {\n readonly traceId: string\n readonly spanId: string\n readonly parentSpanId: string | undefined\n readonly name: string\n readonly startMs: number\n readonly durationMs: number\n readonly ok: boolean\n readonly errorMessage: string | undefined\n}\n\nexport const recordedSpans: RecordedSpan[] = []\n\nexport class TreeSpanExporter {\n export(spans: ReadonlyArray, resultCallback: (result: { code: number }) => void): void {\n for (const span of spans) {\n const [startSec, startNanoRemainder] = span.startTime\n recordedSpans.push({\n traceId: span.spanContext().traceId,\n spanId: span.spanContext().spanId,\n parentSpanId: span.parentSpanContext?.spanId,\n name: span.name,\n startMs: startSec * 1000 + startNanoRemainder / 1e6,\n durationMs: span.duration[0] * 1000 + span.duration[1] / 1e6,\n ok: span.status.code !== 2,\n errorMessage: span.status.message,\n })\n }\n resultCallback({ code: 0 })\n }\n\n shutdown(): Promise {\n return Promise.resolve()\n }\n\n forceFlush(): Promise {\n return Promise.resolve()\n }\n}\n\nexport function printTraceTrees(spans: ReadonlyArray = recordedSpans): void {\n const byTrace = new Map()\n for (const s of spans) {\n const group = byTrace.get(s.traceId) ?? []\n group.push(s)\n byTrace.set(s.traceId, group)\n }\n\n for (const [traceId, group] of byTrace) {\n const byParent = new Map()\n for (const s of group) {\n const siblings = byParent.get(s.parentSpanId) ?? []\n siblings.push(s)\n byParent.set(s.parentSpanId, siblings)\n }\n\n const traceStart = Math.min(...group.map((s) => s.startMs))\n const maxDuration = Math.max(...group.map((s) => s.durationMs))\n const barWidth = 24\n\n console.log(`\\ntrace ${traceId.slice(0, 8)}`)\n\n const printNode = (span: RecordedSpan, prefix: string, isLast: boolean): void => {\n const children = (byParent.get(span.spanId) ?? []).sort((a, b) => a.startMs - b.startMs)\n const connector = isLast ? \"└─ \" : \"├─ \"\n const filled = Math.max(1, Math.round((span.durationMs / maxDuration) * barWidth))\n const bar = \"#\".repeat(filled).padEnd(barWidth, \" \")\n const offset = Math.round(span.startMs - traceStart)\n const status = span.ok ? \"\" : ` ✗ ${span.errorMessage ?? \"failed\"}`\n console.log(\n `${prefix}${connector}${span.name.padEnd(24)} [${bar}] ${Math.round(span.durationMs)}ms @+${offset}ms${status}`,\n )\n const childPrefix = prefix + (isLast ? \" \" : \"│ \")\n children.forEach((child, i) => printNode(child, childPrefix, i === children.length - 1))\n }\n\n const roots = (byParent.get(undefined) ?? []).sort((a, b) => a.startMs - b.startMs)\n roots.forEach((root, i) => printNode(root, \"\", i === roots.length - 1))\n }\n}\n","label",[131],"80cf747e-bdc3-4aad-90bd-f3eef08f8e00","main.ts","2acb14b8-8691-11f1-9e90-1607ee4eb77e",1,["D",1784809347112],"http","import { serveFile } from \"https://esm.town/v/std/utils/index.ts\";\nimport { Data, Duration, Effect, ManagedRuntime, Schedule, Schema } from \"https://esm.sh/effect@3.22.0\";\nimport * as NodeSdk from \"https://esm.sh/@effect/opentelemetry@0.64.0/NodeSdk\";\nimport { ConsoleSpanExporter, SimpleSpanProcessor } from \"https://esm.sh/@opentelemetry/sdk-trace-base@2\";\nimport { runArticleDemo } from \"./articles-demo.ts\";\n\nconst GREENPT_BASE_URL = \"https://api.greenpt.ai/v1\";\nconst GREENPT_TAKEOVER_MODEL = \"green-l-raw\"; // Mistral Small 3.2, fast, non-reasoning\nconst GREENPT_SUGGEST_MODEL = \"qwen3-coder-30b-a3b-instruct\"; // completion-tuned\n\n// Real OTel spans, exported over the same standard pipe Langfuse/any collector already reads\n// (ConsoleSpanExporter + SimpleSpanProcessor here so spans flush per-request for the demo;\n// swap for an OTLP exporter + BatchSpanProcessor to ship elsewhere).\nconst NodeSdkLive = NodeSdk.layer(() => ({\n resource: { serviceName: \"effect-ts-talk\" },\n spanProcessor: new SimpleSpanProcessor(new ConsoleSpanExporter()),\n}));\nconst runtime = ManagedRuntime.make(NodeSdkLive);\n\nclass GreenptApiError extends Data.TaggedError(\"GreenptApiError\")<{ status: number; detail: string }> {}\nclass GreenptTimeout extends Data.TaggedError(\"GreenptTimeout\")<{}> {}\n\nconst CompletionResponse = Schema.Struct({ completion: Schema.String });\nconst decodeCompletion = Schema.decodeUnknown(Schema.parseJson(CompletionResponse));\n\nfunction greenptChat(model: string, system: string, user: string, maxTokens: number) {\n return Effect.gen(function* () {\n const apiKey = Deno.env.get(\"GREENPT_API_KEY\");\n if (!apiKey) return yield* Effect.die(new Error(\"GREENPT_API_KEY is not set\"));\n\n const res = yield* Effect.tryPromise({\n try: (signal) =>\n fetch(`${GREENPT_BASE_URL}/chat/completions`, {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\", Authorization: `Bearer ${apiKey}` },\n body: JSON.stringify({\n model,\n max_tokens: maxTokens,\n temperature: 0.7,\n messages: [\n { role: \"system\", content: system },\n { role: \"user\", content: user },\n ],\n }),\n signal,\n }),\n catch: (cause) => new GreenptApiError({ status: 0, detail: String(cause) }),\n });\n\n if (!res.ok) {\n const detail = yield* Effect.promise(() => res.text());\n return yield* Effect.fail(new GreenptApiError({ status: res.status, detail: detail.slice(0, 300) }));\n }\n\n const raw = yield* Effect.promise(() => res.json());\n const content = raw?.choices?.[0]?.message?.content ?? \"\";\n yield* Effect.annotateCurrentSpan({ contentLen: content.length, usage: raw?.usage });\n return content as string;\n }).pipe(\n Effect.timeoutFail({ duration: Duration.seconds(20), onTimeout: () => new GreenptTimeout() }),\n Effect.retry({\n schedule: Schedule.exponential(Duration.millis(300)).pipe(Schedule.intersect(Schedule.recurs(2))),\n while: (e) => e._tag === \"GreenptTimeout\" || (e._tag === \"GreenptApiError\" && e.status >= 500),\n }),\n Effect.withSpan(\"greenpt.chat\", { attributes: { model, maxTokens } }),\n );\n}\n\nfunction continueText(text: string) {\n const snippet = text.slice(-1500);\n return greenptChat(\n GREENPT_TAKEOVER_MODEL,\n \"You are an AI that hijacks a piece of in-progress writing the moment its human author stops typing. \" +\n \"Continue their text from exactly where it left off, matching their tone, topic, and language. \" +\n \"Write 1-2 short sentences only. Do not repeat what is already there, do not add quotation marks, \" +\n \"labels, or commentary — output only the continuation, starting with a leading space if the text \" +\n \"does not already end with whitespace.\",\n snippet || \"Begin a short piece of reflective writing.\",\n 150,\n ).pipe(Effect.withSpan(\"continueText\"));\n}\n\nfunction buildContext(text: string) {\n const snippet = text.slice(-1200);\n return greenptChat(\n GREENPT_TAKEOVER_MODEL,\n \"You are a context extractor for a writing assistant. Given in-progress text, produce a short \" +\n \"plain-text context block (3-6 sentences max) covering: the topic and tone, any implicit \" +\n \"questions or requests the writer seems to be building toward, and key facts or entities \" +\n \"established so far. Be concise and factual. Do not add opinions or complete the writing.\",\n snippet || \"(empty)\",\n 200,\n ).pipe(Effect.withSpan(\"buildContext\"));\n}\n\nfunction suggestCompletion(text: string) {\n const snippet = text.slice(-800);\n return Effect.gen(function* () {\n const context = yield* buildContext(text);\n\n const systemPrompt =\n 'You are an inline writing completion assistant. Given the tail of an in-progress piece, ' +\n 'return a JSON object {\"completion\":\"...\"} with a short inline completion that finishes the ' +\n \"current sentence without starting a new one — unless there is a clear question or request in \" +\n \"the text that you can answer directly, in which case answer it concisely as the completion. \" +\n 'If nothing is needed, return {\"completion\":\"\"}. Output only valid JSON, no markdown, no code fences.\\n\\n' +\n `\\n${context}\\n`;\n\n const raw = yield* greenptChat(GREENPT_SUGGEST_MODEL, systemPrompt, snippet || \"I\", 100);\n const cleaned = raw.replace(/^```(?:json)?\\s*/i, \"\").replace(/\\s*```$/, \"\").trim();\n\n const { completion } = yield* decodeCompletion(cleaned).pipe(\n Effect.catchTag(\"ParseError\", () => Effect.succeed({ completion: \"\" })),\n );\n return completion;\n }).pipe(Effect.withSpan(\"suggestCompletion\"));\n}\n\nfunction jsonRoute(\n req: Request,\n routeName: string,\n handler: (text: string) => Effect.Effect,\n onError: (error: GreenptApiError | GreenptTimeout) => Response,\n) {\n return Effect.gen(function* () {\n const { text } = yield* Effect.promise(() => req.json());\n const body = yield* handler(typeof text === \"string\" ? text : \"\");\n return Response.json(body);\n }).pipe(\n Effect.withSpan(`http ${routeName}`),\n Effect.catchTags({\n GreenptApiError: (e) => Effect.succeed(onError(e)),\n GreenptTimeout: (e) => Effect.succeed(onError(e)),\n }),\n );\n}\n\nconst upstreamError = (e: GreenptApiError | GreenptTimeout) =>\n Response.json({ error: e._tag === \"GreenptTimeout\" ? \"timeout\" : e.detail }, { status: 502 });\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n\n if (req.method === \"POST\" && url.pathname === \"/ai/continue\") {\n return runtime.runPromise(\n jsonRoute(req, \"/ai/continue\", (text) => continueText(text).pipe(Effect.map((continuation) => ({ continuation }))), upstreamError),\n );\n }\n\n if (req.method === \"POST\" && url.pathname === \"/ai/context\") {\n return runtime.runPromise(\n jsonRoute(req, \"/ai/context\", (text) => buildContext(text).pipe(Effect.map((context) => ({ context }))), upstreamError),\n );\n }\n\n if (req.method === \"POST\" && url.pathname === \"/ai/suggest\") {\n return runtime.runPromise(\n jsonRoute(\n req,\n \"/ai/suggest\",\n (text) => suggestCompletion(text).pipe(Effect.map((completion) => ({ completion }))),\n () => Response.json({ completion: \"\" }), // soft-fail: don't surface to user\n ),\n );\n }\n\n if (req.method === \"POST\" && url.pathname === \"/demo/article\") {\n // Runs on its own Effect v4 runtime (articles-demo.ts); v3 and v4 Effect\n // values can't cross a pipe/provide boundary, only this Promise does.\n try {\n const body = await req.json();\n const article = await runArticleDemo(body);\n return Response.json(article);\n } catch (cause) {\n return Response.json({ error: String(cause) }, { status: 500 });\n }\n }\n\n switch (url.pathname) {\n case \"/style.css\": return serveFile(\"/style.css\", import.meta.url);\n case \"/client.js\": return serveFile(\"/client.js\", import.meta.url);\n case \"/tiny-lm.js\": return serveFile(\"/tiny-lm.js\", import.meta.url);\n case \"/debug.js\": return serveFile(\"/debug.js\", import.meta.url);\n case \"/llm.html\": return serveFile(\"/llm.html\", import.meta.url);\n case \"/llm.js\": return serveFile(\"/llm.js\", import.meta.url);\n case \"/\":\n case \"/index.html\": return serveFile(\"/index.html\", import.meta.url);\n default: return new Response(\"Not found\", { status: 404 });\n }\n}\n",[141],"8a589874-16c2-4aa9-bbd0-9f4281c28b03","main.before.ts","376ccc7a-8691-11f1-911d-1607ee4eb77e",2,["D",1784809368274],"import { serveFile } from \"https://esm.town/v/std/utils/index.ts\";\n\nconst GREENPT_BASE_URL = \"https://api.greenpt.ai/v1\";\nconst GREENPT_TAKEOVER_MODEL = \"green-l-raw\"; // Mistral Small 3.2, fast, non-reasoning\nconst GREENPT_SUGGEST_MODEL = \"qwen3-coder-30b-a3b-instruct\"; // completion-tuned\n\nfunction log(level: \"info\" | \"error\" | \"debug\", msg: string, data?: unknown) {\n const entry = JSON.stringify({ ts: new Date().toISOString(), level, msg, ...(data !== undefined && { data }) });\n if (level === \"error\") console.error(entry);\n else console.log(entry);\n}\n\nasync function greenptChat(model: string, system: string, user: string, maxTokens: number): Promise {\n const apiKey = Deno.env.get(\"GREENPT_API_KEY\");\n if (!apiKey) throw new Error(\"GREENPT_API_KEY is not set\");\n\n log(\"info\", \"greenpt request\", { model, maxTokens });\n const t0 = Date.now();\n const res = await fetch(`${GREENPT_BASE_URL}/chat/completions`, {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\", Authorization: `Bearer ${apiKey}` },\n body: JSON.stringify({\n model,\n max_tokens: maxTokens,\n temperature: 0.7,\n messages: [\n { role: \"system\", content: system },\n { role: \"user\", content: user },\n ],\n }),\n });\n\n if (!res.ok) {\n const detail = await res.text();\n log(\"error\", \"greenpt error\", { model, status: res.status, durationMs: Date.now() - t0, detail: detail.slice(0, 300) });\n throw new Error(`GreenPT ${res.status}: ${detail.slice(0, 300)}`);\n }\n\n const raw = await res.json();\n log(\"debug\", \"greenpt raw\", { model, raw });\n const content = raw?.choices?.[0]?.message?.content ?? \"\";\n log(\"info\", \"greenpt done\", { model, contentLen: content.length, durationMs: Date.now() - t0, usage: raw?.usage });\n return content;\n}\n\nasync function continueText(text: string): Promise {\n const snippet = text.slice(-1500);\n return greenptChat(\n GREENPT_TAKEOVER_MODEL,\n \"You are an AI that hijacks a piece of in-progress writing the moment its human author stops typing. \" +\n \"Continue their text from exactly where it left off, matching their tone, topic, and language. \" +\n \"Write 1-2 short sentences only. Do not repeat what is already there, do not add quotation marks, \" +\n \"labels, or commentary — output only the continuation, starting with a leading space if the text \" +\n \"does not already end with whitespace.\",\n snippet || \"Begin a short piece of reflective writing.\",\n 150,\n );\n}\n\nasync function buildContext(text: string): Promise {\n const snippet = text.slice(-1200);\n return greenptChat(\n GREENPT_TAKEOVER_MODEL,\n \"You are a context extractor for a writing assistant. Given in-progress text, produce a short \" +\n \"plain-text context block (3-6 sentences max) covering: the topic and tone, any implicit \" +\n \"questions or requests the writer seems to be building toward, and key facts or entities \" +\n \"established so far. Be concise and factual. Do not add opinions or complete the writing.\",\n snippet || \"(empty)\",\n 200,\n );\n}\n\nasync function suggestCompletion(text: string): Promise {\n const snippet = text.slice(-800);\n\n const context = await buildContext(text);\n log(\"debug\", \"suggestCompletion context\", { context });\n\n const systemPrompt =\n 'You are an inline writing completion assistant. Given the tail of an in-progress piece, ' +\n 'return a JSON object {\"completion\":\"...\"} with a short inline completion that finishes the ' +\n \"current sentence without starting a new one — unless there is a clear question or request in \" +\n \"the text that you can answer directly, in which case answer it concisely as the completion. \" +\n 'If nothing is needed, return {\"completion\":\"\"}. Output only valid JSON, no markdown, no code fences.\\n\\n' +\n `\\n${context}\\n`;\n\n const raw = await greenptChat(GREENPT_SUGGEST_MODEL, systemPrompt, snippet || \"I\", 100);\n\n try {\n const cleaned = raw.replace(/^```(?:json)?\\s*/i, \"\").replace(/\\s*```$/, \"\").trim();\n const parsed = JSON.parse(cleaned);\n return typeof parsed.completion === \"string\" ? parsed.completion : \"\";\n } catch {\n log(\"error\", \"suggestCompletion parse failed\", { raw });\n return \"\";\n }\n}\n\nexport default async function (req: Request): Promise {\n const url = new URL(req.url);\n\n if (req.method === \"POST\" && url.pathname === \"/ai/continue\") {\n const t0 = Date.now();\n try {\n const { text } = await req.json();\n log(\"info\", \"POST /ai/continue\", { textLen: typeof text === \"string\" ? text.length : null });\n const continuation = await continueText(typeof text === \"string\" ? text : \"\");\n log(\"info\", \"POST /ai/continue done\", { durationMs: Date.now() - t0 });\n return Response.json({ continuation });\n } catch (err) {\n log(\"error\", \"POST /ai/continue failed\", { error: String(err), durationMs: Date.now() - t0 });\n return Response.json({ error: String(err) }, { status: 500 });\n }\n }\n\n if (req.method === \"POST\" && url.pathname === \"/ai/context\") {\n const t0 = Date.now();\n try {\n const { text } = await req.json();\n log(\"info\", \"POST /ai/context\", { textLen: typeof text === \"string\" ? text.length : null });\n const context = await buildContext(typeof text === \"string\" ? text : \"\");\n log(\"info\", \"POST /ai/context done\", { durationMs: Date.now() - t0 });\n return Response.json({ context });\n } catch (err) {\n log(\"error\", \"POST /ai/context failed\", { error: String(err), durationMs: Date.now() - t0 });\n return Response.json({ error: String(err) }, { status: 500 });\n }\n }\n\n if (req.method === \"POST\" && url.pathname === \"/ai/suggest\") {\n const t0 = Date.now();\n try {\n const { text } = await req.json();\n log(\"info\", \"POST /ai/suggest\", { textLen: typeof text === \"string\" ? text.length : null });\n const completion = await suggestCompletion(typeof text === \"string\" ? text : \"\");\n log(\"info\", \"POST /ai/suggest done\", { durationMs: Date.now() - t0 });\n return Response.json({ completion });\n } catch (err) {\n log(\"error\", \"POST /ai/suggest failed\", { error: String(err), durationMs: Date.now() - t0 });\n return Response.json({ completion: \"\" }); // soft-fail: don't surface to user\n }\n }\n\n switch (url.pathname) {\n case \"/style.css\": return serveFile(\"/style.css\", import.meta.url);\n case \"/client.js\": return serveFile(\"/client.js\", import.meta.url);\n case \"/tiny-lm.js\": return serveFile(\"/tiny-lm.js\", import.meta.url);\n case \"/debug.js\": return serveFile(\"/debug.js\", import.meta.url);\n case \"/llm.html\": return serveFile(\"/llm.html\", import.meta.url);\n case \"/llm.js\": return serveFile(\"/llm.js\", import.meta.url);\n case \"/\":\n case \"/index.html\": return serveFile(\"/index.html\", import.meta.url);\n default: return new Response(\"Not found\", { status: 404 });\n }\n}\n",[149],"68117cdd-8f7f-4988-be3b-ebc18f02e13f","articles-effect.ts","418e886a-8691-11f1-9d1e-1607ee4eb77e",3,["D",1784809385337],"/**\n * Effect version of `createArticle` — line-for-line against the OTel/async\n * original. Same behavior: root span, attributes, fire-and-forget\n * notification that must not fail the request, error recorded on the span.\n *\n * What does NOT exist in this file, because the runtime owns it:\n * - tracer.startActiveSpan / span.end / finally (span = fiber scope)\n * - span.setStatus(OK|ERROR) / span.recordException (Exit → span status)\n * - try/catch + `throw error` re-raise (typed error channel)\n * - `.catch((err) => console.error(...))` (fork + tapCause)\n * - `error as Error` casts (errors are values)\n *\n * Requires Effect v4 (Context.Service): pinned below to the v4 beta that\n * matches @effect/opentelemetry's v4 beta tag used in articles-demo.ts.\n */\nimport {\n Context,\n Data,\n Effect,\n Layer,\n} from \"https://esm.sh/effect@4.0.0-beta.101\";\n\n// --- Dependencies as services (stand-ins for drizzle + the job queue) -------\n\nexport class ArticleInsertError extends Data.TaggedError(\"ArticleInsertError\")<{\n readonly slug: string;\n readonly cause: unknown;\n}> {}\n\ninterface ArticleRow {\n readonly id: number;\n readonly slug: string;\n readonly title: string;\n readonly authorId: number;\n}\n\nexport class Database extends Context.Service<\n Database,\n {\n readonly insertArticle: (values: {\n readonly slug: string;\n readonly title: string;\n readonly description: string | null;\n readonly body: string;\n readonly authorId: number;\n }) => Effect.Effect;\n }\n>()(\"app/Database\") {}\n\nexport class Notifications extends Context.Service<\n Notifications,\n {\n readonly articleCreated: (input: {\n readonly articleId: number;\n readonly articleSlug: string;\n readonly authorId: number;\n readonly title: string;\n }) => Effect.Effect;\n }\n>()(\"app/Notifications\") {}\n\nconst generateSlug = (title: string) =>\n title.toLowerCase().replace(/[^a-z0-9]+/g, \"-\").replace(/^-|-$/g, \"\");\n\nexport class Articles extends Context.Service()(\"app/Articles\", {\n make: Effect.gen(function* () {\n const db = yield* Database;\n const notifications = yield* Notifications;\n\n const create = Effect.fn(\"article.create\")(\n function* (\n authorId: number,\n input: { title: string; description?: string; body: string },\n ) {\n const slug = generateSlug(input.title);\n\n const article = yield* db.insertArticle({\n slug,\n title: input.title,\n description: input.description ?? null,\n body: input.body,\n authorId,\n });\n\n yield* Effect.annotateCurrentSpan({\n \"user.id\": authorId,\n \"article.id\": article.id,\n });\n\n // Fire-and-forget, but NOT invisible: the forked fiber inherits the\n // trace, gets its own child span, and failures land on that span's\n // exit + the logs — instead of a floating promise with .catch(console.error).\n yield* notifications\n .articleCreated({\n articleId: article.id,\n articleSlug: article.slug,\n authorId,\n title: article.title,\n })\n .pipe(\n Effect.withSpan(\"notification.articleCreated\"),\n Effect.tapCause((cause) =>\n Effect.logWarning(\"Failed to enqueue notification\", cause)\n ),\n Effect.forkDetach,\n );\n\n return article;\n },\n );\n\n return { create } as const;\n }),\n}) {\n static readonly layerNoDeps = Layer.effect(this, this.make);\n}\n\n// --- Tenant attribution: the multi-tenant / single-tenant switch -------------\n//\n// In the OTel version, tenant attributes on every span require a custom\n// SpanProcessor wired into the SDK, or setAttribute calls sprinkled into\n// every service. Here it's `Effect.annotateSpans` — a combinator applied at\n// whatever scope owns the tenancy decision. The service code above never\n// changes.\n\n// Single-tenant deployment: annotate at the composition root, once, at boot.\n// Every span in the entire app carries the attributes.\nexport const annotateDeployment = (app: Effect.Effect) =>\n Effect.annotateSpans(app, {\n \"app.tenant_id\": \"org_school_001\",\n \"app.data_region\": \"eu-west\",\n });\n\n// Multi-tenant api-proxy: annotate at the request boundary. Every span under\n// this request — including the forked notification fiber, which inherits the\n// annotations — carries the authenticated tenant.\nexport const handleRequest = (\n tenant: { id: string; region: string },\n authorId: number,\n input: { title: string; body: string },\n) =>\n Effect.gen(function* () {\n const articles = yield* Articles;\n return yield* articles.create(authorId, input);\n }).pipe(\n Effect.annotateSpans({\n \"app.tenant_id\": tenant.id,\n \"app.data_region\": tenant.region,\n }),\n );\n",[156],"be1150e6-b9b2-4bad-8495-e62bf5223e08",["D",1784809411130],[27],"pulls","branches","forks","envVars",{"_172":59,"_174":59},"connections",{"_172":59,"_173":59},"count","unfilledCount","emptyCount",{"_41":162,"_31":27,"_29":30,"_133":-5,"_39":40,"_38":14,"_135":120,"_36":193,"_13":59,"_32":33,"_34":35,"_23":24,"_138":-5,"_181":194,"_26":195},{"_41":155,"_31":156,"_29":157,"_133":-5,"_39":40,"_38":158,"_135":120,"_36":190,"_13":59,"_32":33,"_34":160,"_23":24,"_138":-5,"_181":191,"_26":192},{"_41":148,"_31":149,"_29":150,"_133":-5,"_39":40,"_38":151,"_135":24,"_36":187,"_13":59,"_32":145,"_34":153,"_23":24,"_138":-5,"_181":188,"_26":189},{"_41":140,"_31":141,"_29":142,"_133":-5,"_39":40,"_38":143,"_135":24,"_36":184,"_13":59,"_32":145,"_34":146,"_23":24,"_138":-5,"_181":185,"_26":186},{"_41":130,"_31":131,"_29":132,"_133":-5,"_39":40,"_38":134,"_135":120,"_36":180,"_13":59,"_32":33,"_34":137,"_23":24,"_138":-5,"_181":182,"_26":183},["D",1784809399012],"children",[],[131],["D",1784809347112],[],[141],["D",1784809368274],[],[149],["D",1784809385337],[],[156],["D",1784809411130],[],[27],["D",1784809347061],{"_93":92,"_101":97},{"_93":92,"_94":95,"_96":97,"_18":53,"_98":99,"_100":24},{"_13":82,"_31":83,"_41":40,"_84":42,"_85":-5,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":43,"_91":92},"ctx",{"_75":-5,"_122":233,"_234":235,"_236":237,"_238":24,"_239":24,"_240":24,"_241":120,"_242":-5,"_243":-7,"_244":-7,"_245":59,"_246":24,"_247":24,"_248":-5,"_249":24,"_250":24,"_251":24,"_252":-5},"ENV",{"_204":205,"_206":207,"_208":83,"_209":210,"_211":212,"_213":214,"_215":216,"_217":218,"_219":220,"_221":222,"_223":224,"_225":226,"_227":228,"_229":230,"_231":232},"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 · v5