alys-akusa 0.1.6 → 0.1.7

package/dist/index.cjs

@@ -6477,22 +6477,22 @@ async function generateWithOpenAI(options) {
   };
 }
 async function generateWithOpenAIBatched(options) {
-  const batchSize = Math.max(1, Math.min(24, Number(process.env.ALYS_OPENAI_RECORDS_PER_CALL ?? 8)));
-  const batches = Math.ceil(options.targetCount / batchSize);
-  const records = [];
-  let latencyMs = 0;
-  let model = process.env.ALYS_GENERATOR_MODEL || DEFAULT_OPENAI_MODEL;
-  for (let i = 0; i < batches; i++) {
-    const remaining = options.targetCount - records.length;
-    if (remaining <= 0) break;
-    const result = await generateWithOpenAI({
+  const batchSize = Math.max(1, Math.min(48, Number(process.env.ALYS_OPENAI_RECORDS_PER_CALL ?? 8)));
+  const batches = Array.from(
+    { length: Math.ceil(options.targetCount / batchSize) },
+    (_, index) => Math.min(batchSize, Math.max(0, options.targetCount - index * batchSize))
+  ).filter((count) => count > 0);
+  const results = await mapLimit(
+    batches,
+    Number(process.env.ALYS_PROVIDER_BATCH_CONCURRENCY ?? 3),
+    (count) => generateWithOpenAI({
       ...options,
-      targetCount: Math.min(batchSize, remaining)
-    });
-    latencyMs += result.latencyMs;
-    model = result.model;
-    records.push(...result.records);
-  }
+      targetCount: count
+    })
+  );
+  const records = results.flatMap((result) => result.records);
+  const latencyMs = results.reduce((sum, result) => sum + result.latencyMs, 0);
+  const model = results[0]?.model ?? process.env.ALYS_GENERATOR_MODEL ?? DEFAULT_OPENAI_MODEL;
   return {
     provider: "openai",
     model,
@@ -6543,22 +6543,22 @@ async function generateWithGroq(options) {
   };
 }
 async function generateWithGroqBatched(options) {
-  const batchSize = Math.max(1, Math.min(16, Number(process.env.ALYS_GROQ_RECORDS_PER_CALL ?? 8)));
-  const batches = Math.ceil(options.targetCount / batchSize);
-  const records = [];
-  let latencyMs = 0;
-  let model = process.env.GROQ_MODEL || process.env.ALYS_GENERATOR_MODEL || DEFAULT_GROQ_MODEL;
-  for (let i = 0; i < batches; i++) {
-    const remaining = options.targetCount - records.length;
-    if (remaining <= 0) break;
-    const result = await generateWithGroq({
+  const batchSize = Math.max(1, Math.min(32, Number(process.env.ALYS_GROQ_RECORDS_PER_CALL ?? 8)));
+  const batches = Array.from(
+    { length: Math.ceil(options.targetCount / batchSize) },
+    (_, index) => Math.min(batchSize, Math.max(0, options.targetCount - index * batchSize))
+  ).filter((count) => count > 0);
+  const results = await mapLimit(
+    batches,
+    Number(process.env.ALYS_PROVIDER_BATCH_CONCURRENCY ?? 3),
+    (count) => generateWithGroq({
       ...options,
-      targetCount: Math.min(batchSize, remaining)
-    });
-    latencyMs += result.latencyMs;
-    model = result.model;
-    records.push(...result.records);
-  }
+      targetCount: count
+    })
+  );
+  const records = results.flatMap((result) => result.records);
+  const latencyMs = results.reduce((sum, result) => sum + result.latencyMs, 0);
+  const model = results[0]?.model ?? process.env.GROQ_MODEL ?? process.env.ALYS_GENERATOR_MODEL ?? DEFAULT_GROQ_MODEL;
   return {
     provider: "groq",
     model,
@@ -6566,6 +6566,20 @@ async function generateWithGroqBatched(options) {
     records: records.slice(0, options.targetCount)
   };
 }
+async function mapLimit(items, limit, worker) {
+  const results = new Array(items.length);
+  let nextIndex = 0;
+  const workerCount = Math.max(1, Math.min(items.length, Math.floor(limit || 1)));
+  async function runWorker() {
+    while (nextIndex < items.length) {
+      const index = nextIndex;
+      nextIndex += 1;
+      results[index] = await worker(items[index], index);
+    }
+  }
+  await Promise.all(Array.from({ length: workerCount }, runWorker));
+  return results;
+}
 function providerRecordSchema() {
   return {
     type: "object",
@@ -6680,7 +6694,7 @@ var DiscoveryAgent = class {
 var ExtractionAgent = class {
   name = "ExtractionAgent";
   async run(sources) {
-    const crawls = await mapLimit(
+    const crawls = await mapLimit2(
       sources,
       Math.max(1, Number(process.env.ALYS_CRAWL_CONCURRENCY ?? 6)),
       (source) => crawlSource(source)
@@ -6754,7 +6768,7 @@ var StructuringAgent = class {
     }
     const concurrency = Math.max(1, Math.min(documents.length, Number(process.env.ALYS_PROVIDER_CONCURRENCY ?? 3)));
     let totalGenerated = 0;
-    const grouped = await mapLimit(documents, concurrency, async (document, index) => {
+    const grouped = await mapLimit2(documents, concurrency, async (document, index) => {
       const finding = findings[index] ?? findings[0];
       const trustScore = document.sourceScores?.trustScore ?? 0.62;
       const authorityScore = document.sourceScores?.authorityScore ?? 0.55;
@@ -6830,7 +6844,7 @@ function domainFromUrl3(url) {
     return "unknown";
   }
 }
-async function mapLimit(items, limit, worker) {
+async function mapLimit2(items, limit, worker) {
   const results = new Array(items.length);
   let nextIndex = 0;
   async function runWorker() {
@@ -7313,7 +7327,7 @@ async function generateDatasets(options) {
   const verificationEnabled = options.enableVerificationSwarm ?? performanceMode !== "fast";
   const debateEnabled = verificationEnabled && perf.debateEnabled;
   const multiplier = depthMultiplier(options.generationDepth);
-  const results = await mapLimit2(Array.from({ length: datasetCount }, (_, i) => i), Number(process.env.ALYS_DATASET_CONCURRENCY ?? perf.datasetConcurrency), async (i) => {
+  const results = await mapLimit3(Array.from({ length: datasetCount }, (_, i) => i), Number(process.env.ALYS_DATASET_CONCURRENCY ?? perf.datasetConcurrency), async (i) => {
     const datasetIndex = i + 1;
     const datasetSourceLimit = Math.max(1, Math.floor((options.sourceLimit ?? 24) * multiplier));
     const datasetTargetRows = Math.max(1, Math.floor((options.targetRows ?? 100) * multiplier));
@@ -7343,7 +7357,7 @@ async function generateDatasets(options) {
   const datasets = results;
   return { manifests, artifacts, previews, datasets };
 }
-async function mapLimit2(items, limit, worker) {
+async function mapLimit3(items, limit, worker) {
   const results = new Array(items.length);
   let nextIndex = 0;
   const workerCount = Math.max(1, Math.min(items.length, Math.floor(limit || 1)));
@@ -7682,7 +7696,7 @@ function printUsage(profile) {
   );
 }
 function printRunPlan(args) {
-  const multiplier = depthMultiplier2(args.depth);
+  const multiplier = args.performanceMode === "fast" ? 1 : depthMultiplier2(args.depth);
   const effectiveSources = Math.max(1, Math.floor(args.sourceLimit * multiplier));
   const effectiveRows = Math.max(1, Math.floor(args.targetRows * multiplier));
   const totalRows = effectiveRows * args.datasetCount;
@@ -8000,7 +8014,7 @@ async function handleGenerate(args, command) {
     validate: (v) => v.trim().length ? true : "Please enter a topic."
   })).topic);
   if (!topic) throw new Error("Missing topic.");
-  const datasetType = parseDatasetType(values.type) ?? (await (0, import_prompts3.default)({
+  const datasetType = parseDatasetType(values.type) ?? (values.yes === true ? "instruction" : void 0) ?? (await (0, import_prompts3.default)({
     type: "select",
     name: "datasetType",
     message: "Dataset type?",
@@ -8010,7 +8024,7 @@ async function handleGenerate(args, command) {
       { title: "Question/Answer", value: "qa" }
     ]
   })).datasetType;
-  const requestedDatasetCount = values.datasets ? Math.max(1, Math.floor(Number(values.datasets))) : (await (0, import_prompts3.default)({
+  const requestedDatasetCount = values.datasets ? Math.max(1, Math.floor(Number(values.datasets))) : values.yes === true ? 1 : (await (0, import_prompts3.default)({
     type: "number",
     name: "datasetCount",
     message: "How many datasets?",
@@ -8023,7 +8037,7 @@ async function handleGenerate(args, command) {
     return;
   }
   const datasetCount = requestedDatasetCount;
-  const exportFormats = values.format ? parseFormats(values.format) : (await (0, import_prompts3.default)({
+  const exportFormats = values.format ? parseFormats(values.format) : values.yes === true ? ["jsonl", "csv", "markdown"] : (await (0, import_prompts3.default)({
     type: "multiselect",
     name: "exportFormats",
     message: "Output formats?",
@@ -8036,7 +8050,7 @@ async function handleGenerate(args, command) {
     ],
     hint: "Use space to select multiple."
   })).exportFormats;
-  const depth = parseDepth(values.depth) ?? (await (0, import_prompts3.default)({
+  const depth = parseDepth(values.depth) ?? (values.yes === true ? "medium" : void 0) ?? (await (0, import_prompts3.default)({
     type: "select",
     name: "depth",
     message: "Research depth?",
@@ -8046,7 +8060,7 @@ async function handleGenerate(args, command) {
       { title: "Deep", value: "deep" }
     ]
   })).depth;
-  const sourceLimit = values.sources ? Math.min(maxSources, Math.max(1, Number(values.sources))) : (await (0, import_prompts3.default)({
+  const sourceLimit = values.sources ? Math.min(maxSources, Math.max(1, Number(values.sources))) : values.yes === true ? benchmarkMode ? 48 : MAX_SOURCES_PER_RUN : (await (0, import_prompts3.default)({
     type: "number",
     name: "sourceLimit",
     message: "How many sources?",
@@ -8054,7 +8068,7 @@ async function handleGenerate(args, command) {
     min: 1,
     max: maxSources
   })).sourceLimit;
-  const targetRows = values.rows ? Math.min(maxRows, Math.max(1, Number(values.rows))) : (await (0, import_prompts3.default)({
+  const targetRows = values.rows ? Math.min(maxRows, Math.max(1, Number(values.rows))) : values.yes === true ? benchmarkMode ? 5e3 : MAX_ROWS_PER_DATASET : (await (0, import_prompts3.default)({
     type: "number",
     name: "targetRows",
     message: "Rows per dataset?",
@@ -8062,14 +8076,14 @@ async function handleGenerate(args, command) {
     min: 1,
     max: maxRows
   })).targetRows;
-  const workspaceRoot = (values.workspace ? String(values.workspace) : "").trim() || (await (0, import_prompts3.default)({
+  const workspaceRoot = (values.workspace ? String(values.workspace) : "").trim() || (values.yes === true ? "~/Alys" : "") || (await (0, import_prompts3.default)({
     type: "text",
     name: "workspaceRoot",
     message: "Export directory?",
     initial: "~/Alys",
     validate: (v) => v.trim().length ? true : "Enter an export directory."
   })).workspaceRoot;
-  const verificationEnabled = values.verify === true ? true : values["no-verify"] === true ? false : (await (0, import_prompts3.default)({
+  const verificationEnabled = values.verify === true ? true : values["no-verify"] === true ? false : values.yes === true ? performanceMode !== "fast" : (await (0, import_prompts3.default)({
     type: "toggle",
     name: "verificationEnabled",
     message: "Enable verification checks?",
@@ -8119,9 +8133,9 @@ async function handleGenerate(args, command) {
   console.log(paint("Runtime", "white"));
   printStage("AUTH", "OK", "Usage linked", appUrl());
   printStage("PLAN", "OK", "Generations charged only after successful completion", `${datasetCount} requested`);
-  printStage("SRC", "RUN", "Research pipeline starting", `${sourceLimit} source target`);
+  printStage("RUN", "RUN", "Dataset runtime starting", `${performanceMode} mode`);
   const response = await withSpinner(
-    "Alys research runtime executing",
+    "Alys runtime executing",
     requestJson(
       "/api/cli/generate",
       {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "alys-akusa",
-  "version": "0.1.6",
+  "version": "0.1.7",
   "private": false,
   "description": "Alys local CLI runtime for autonomous dataset generation.",
   "license": "UNLICENSED",