alys-akusa 0.1.2 → 0.1.4

package/README.md

@@ -1,6 +1,6 @@
 # Alys CLI
 
-Alys is the authenticated CLI for Akusa-managed dataset generation.
+Alys is the authenticated CLI for terminal-native dataset generation.
 
 ```bash
 npm install -g alys-akusa

package/dist/index.cjs

@@ -4908,6 +4908,8 @@ var CONFIG_DIR = import_node_path.default.join(import_node_os.default.homedir(),
 var CONFIG_PATH = import_node_path.default.join(CONFIG_DIR, "config.json");
 var ALYS_APP_URL = process.env.ALYS_APP_URL || process.env.NEXT_PUBLIC_SITE_URL || "https://alys.akusa.dev";
 var MAX_DATASETS_PER_RUN = 5;
+var MAX_SOURCES_PER_RUN = 8;
+var MAX_ROWS_PER_DATASET = 125;
 var FIGLET_LOGO = [
   "     o      ooooo    ooooo  oooo oooooooo8  ",
   "    888      888       888  88  888         ",
@@ -4928,6 +4930,8 @@ function printBanner() {
 function printHelp() {
   printBanner();
   console.log(`
+Alys runs in your terminal. The dashboard tracks what it generates.
+
 Usage:
   npx alys-akusa
   npx alys-akusa login
@@ -4940,8 +4944,8 @@ Flags:
   --type instruction|rag|qa
   --datasets 1
   --depth shallow|medium|deep
-  --sources 12
-  --rows 1000
+  --sources 8
+  --rows 125
   --workspace ~/Desktop/alys-output
   --verify
   --no-verify
@@ -5014,7 +5018,7 @@ async function requestJson(pathname, init = {}, token) {
   const response = await fetch(apiUrl(pathname), { ...init, headers });
   const payload = await response.json().catch(() => ({}));
   if (!response.ok) {
-    const message = typeof payload?.error === "string" ? payload.error : `Alys API request failed (${response.status}).`;
+    const message = typeof payload?.error === "string" ? payload.error : `Alys request failed (${response.status}).`;
     throw new CliApiError(message, response.status, payload);
   }
   return payload;
@@ -5078,6 +5082,19 @@ function parseDepth(value) {
   if (value === "shallow" || value === "medium" || value === "deep") return value;
   return void 0;
 }
+function normalizeTopicInput(value) {
+  if (typeof value !== "string") return "";
+  let topic = value.replace(/\[[^\]]+\][^\n\r]*/g, "").replace(/\s+/g, " ").trim();
+  const commandMatch = topic.match(/^(?:npx\s+)?(?:alys-akusa(?:@latest)?|alys)\s+generate\s+(.+)$/i);
+  if (commandMatch) {
+    topic = commandMatch[1].trim();
+  }
+  const quoted = topic.match(/^"([^"]+)"|^'([^']+)'/);
+  if (quoted) {
+    return (quoted[1] || quoted[2] || "").trim();
+  }
+  return topic.replace(/\s+--(?:format|type|datasets|depth|sources|rows|workspace)\s+\S.*$/i, "").replace(/\s+--(?:verify|no-verify|yes)\b.*$/i, "").trim();
+}
 function resolveHome(p) {
   if (p.startsWith("~/")) return import_node_path.default.join(import_node_os.default.homedir(), p.slice(2));
   return p;
@@ -5085,14 +5102,94 @@ function resolveHome(p) {
 function formatInt(value) {
   return new Intl.NumberFormat("en-US").format(value);
 }
+function formatPercent(value) {
+  if (!Number.isFinite(value) || value <= 0) return "pending";
+  return `${Math.round(value * 100)}%`;
+}
+function formatScore(value) {
+  if (!Number.isFinite(value) || value <= 0) return "pending";
+  return `${Math.round(value)}/100`;
+}
+function average(values) {
+  const filtered = values.filter((value) => Number.isFinite(value) && value > 0);
+  return filtered.length ? filtered.reduce((sum, value) => sum + value, 0) / filtered.length : 0;
+}
+function truncate(value, max = 88) {
+  const normalized = value.replace(/\s+/g, " ").trim();
+  return normalized.length > max ? `${normalized.slice(0, max - 1)}\u2026` : normalized;
+}
+function getMetrics(dataset) {
+  const metrics = dataset.manifest.metrics;
+  return metrics && typeof metrics === "object" ? metrics : {};
+}
+function getSummary(dataset) {
+  const summary = dataset.manifest.generationSummary;
+  return summary && typeof summary === "object" ? summary : {};
+}
+function getQualityMetrics(dataset) {
+  const metrics = dataset.manifest.qualityMetrics;
+  return metrics && typeof metrics === "object" ? metrics : {};
+}
+function getEvaluation(dataset) {
+  const evaluation = dataset.manifest.evaluation;
+  return evaluation && typeof evaluation === "object" ? evaluation : {};
+}
+function printStage(code, status, label, metric) {
+  const tint = status === "DONE" || status === "OK" ? "green" : status === "WARN" ? "yellow" : "cyan";
+  const prefix = `${paint(`[${code.padEnd(4).slice(0, 4)}]`, "gray")} ${paint(status.padEnd(4), tint)}`;
+  console.log(`${prefix} ${label}${metric ? ` ${paint(metric, "gray")}` : ""}`);
+}
+async function withSpinner(label, task) {
+  if (!process.stdout.isTTY) {
+    printStage("RUN", "RUN", label);
+    return task;
+  }
+  const frames = ["\u280B", "\u2819", "\u2839", "\u2838", "\u283C", "\u2834", "\u2826", "\u2827", "\u2807", "\u280F"];
+  let index = 0;
+  process.stdout.write(`${paint(frames[index], "yellow")} ${label}`);
+  const interval = setInterval(() => {
+    index = (index + 1) % frames.length;
+    process.stdout.write(`\r${paint(frames[index], "yellow")} ${label}`);
+  }, 90);
+  try {
+    const result = await task;
+    clearInterval(interval);
+    process.stdout.write(`\r${paint("\u2713", "green")} ${label}
+`);
+    return result;
+  } catch (error) {
+    clearInterval(interval);
+    process.stdout.write(`\r${paint("\xD7", "red")} ${label}
+`);
+    throw error;
+  }
+}
+function previewRecord(dataset) {
+  const file = dataset.files.find((item) => item.format === "jsonl" || item.format === "instruction" || item.format === "rag");
+  const firstLine = file?.content.split(/\r?\n/).find((line) => line.trim().length > 0);
+  if (!firstLine) return null;
+  try {
+    const parsed = JSON.parse(firstLine);
+    const input = typeof parsed.input === "string" ? parsed.input : typeof parsed.instruction === "string" ? parsed.instruction : typeof parsed.question === "string" ? parsed.question : "";
+    const output = typeof parsed.output === "string" ? parsed.output : typeof parsed.answer === "string" ? parsed.answer : typeof parsed.completion === "string" ? parsed.completion : typeof parsed.text === "string" ? parsed.text : "";
+    if (!input && !output) return null;
+    const metadata = parsed.metadata && typeof parsed.metadata === "object" ? parsed.metadata : {};
+    const explanation = Array.isArray(metadata.acceptance_explanation) ? metadata.acceptance_explanation.filter((item) => typeof item === "string") : Array.isArray(metadata.acceptance_reasons) ? metadata.acceptance_reasons.filter((item) => typeof item === "string") : [];
+    return { input: truncate(input, 92), output: truncate(output, 120), why: explanation.slice(0, 2).map((item) => truncate(item, 112)) };
+  } catch {
+    return null;
+  }
+}
 function depthMultiplier(depth) {
   if (depth === "deep") return 1.6;
   if (depth === "shallow") return 0.75;
   return 1;
 }
 function printUsage(profile) {
-  const name = profile.user.name || profile.user.email || "Alys user";
-  console.log(paint(`Signed in as ${name}`, "white"));
+  const name = profile.user.name || "Alys user";
+  const email = profile.user.email;
+  const account = email && email !== name ? `${name} <${email}>` : name;
+  console.log(paint(`Account: ${account}`, "white"));
   console.log(
     profile.usage.isUnlimited ? paint("Generations: unlimited", "green") : paint(`Generations: ${profile.usage.used}/${profile.usage.limit} used, ${profile.usage.remaining} remaining`, "yellow")
   );
@@ -5102,10 +5199,14 @@ function printRunPlan(args) {
   const effectiveSources = Math.max(1, Math.floor(args.sourceLimit * multiplier));
   const effectiveRows = Math.max(1, Math.floor(args.targetRows * multiplier));
   const totalRows = effectiveRows * args.datasetCount;
-  console.log(paint("\n\u250C\u2500 Alys API run plan \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510", "gray"));
+  const name = args.profile.user.name || "Alys user";
+  const email = args.profile.user.email;
+  const account = email && email !== name ? `${name} <${email}>` : name;
+  console.log(paint("\n\u250C\u2500 Alys run plan \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510", "gray"));
   console.log(`${paint("\u2502", "gray")} Topic        ${paint(args.topic.slice(0, 58), "white")}`);
-  console.log(`${paint("\u2502", "gray")} Type         ${paint(args.datasetType, "cyan")}   Depth ${paint(args.depth, "cyan")}   Mode ${paint("Akusa API", "yellow")}`);
-  console.log(`${paint("\u2502", "gray")} Datasets     ${formatInt(args.datasetCount)} / ${MAX_DATASETS_PER_RUN} max   Rows/dataset ${formatInt(effectiveRows)}`);
+  console.log(`${paint("\u2502", "gray")} Account      ${paint(account.slice(0, 58), "white")}`);
+  console.log(`${paint("\u2502", "gray")} Type         ${paint(args.datasetType, "cyan")}   Depth ${paint(args.depth, "cyan")}   Mode ${paint("Alys runtime", "yellow")}`);
+  console.log(`${paint("\u2502", "gray")} Datasets     ${formatInt(args.datasetCount)} / ${MAX_DATASETS_PER_RUN} max   Rows/dataset ${formatInt(effectiveRows)} max`);
   console.log(`${paint("\u2502", "gray")} Total rows   ${paint(formatInt(totalRows), "green")}   Formats ${args.exportFormats.join(", ")}`);
   console.log(`${paint("\u2502", "gray")} Credits      ${args.profile.usage.isUnlimited ? "Unlimited" : `${args.profile.usage.remaining ?? 0} remaining before run`}`);
   console.log(`${paint("\u2502", "gray")} Output       ${resolveHome(args.workspaceRoot)}`);
@@ -5121,7 +5222,7 @@ async function confirmRun(args) {
   const response = await (0, import_prompts.default)({
     type: "toggle",
     name: "continueRun",
-    message: "Generate through the Alys API?",
+    message: "Start this Alys run?",
     initial: true,
     active: "Yes",
     inactive: "No"
@@ -5144,6 +5245,108 @@ async function writeGeneratedDatasets(workspaceRoot, datasets) {
     );
   }
 }
+function printGenerationSummary(response, workspaceRoot) {
+  const root = resolveHome(workspaceRoot);
+  const totals = response.datasets.reduce(
+    (acc, dataset) => {
+      const metrics = getMetrics(dataset);
+      const summary = getSummary(dataset);
+      acc.records += Number(metrics.recordsGenerated ?? summary.recordsAccepted ?? 0);
+      acc.sources += Number(metrics.sourcesDiscovered ?? 0);
+      acc.documents += Number(metrics.documentsExtracted ?? 0);
+      acc.findings += Number(metrics.findingsVerified ?? 0);
+      acc.duplicates += Number(metrics.duplicatesRemoved ?? summary.duplicatesRemoved ?? 0);
+      const quality = getQualityMetrics(dataset);
+      acc.contradictions += Number(quality.contradictionResolutionCount ?? 0);
+      acc.lowTrustFiltered += Number(quality.lowTrustSourceFilterRate ?? 0);
+      acc.citationCoverage.push(Number(quality.citationCoverage ?? 0));
+      acc.uniqueness.push(Number(quality.recordUniqueness ?? 0));
+      acc.relevance.push(Number(quality.relevanceScore ?? 0));
+      const suitability = getEvaluation(dataset).suitability ?? getEvaluation(dataset).summary?.scores ?? {};
+      acc.ragSuitability.push(Number(suitability.ragSuitability ?? 0));
+      acc.instructionTuning.push(Number(suitability.instructionTuning ?? 0));
+      acc.factualGrounding.push(Number(suitability.factualGrounding ?? 0));
+      acc.humanUsefulness.push(Number(suitability.humanUsefulness ?? 0));
+      const confidence2 = Number(metrics.averageConfidence ?? summary.averageConfidence ?? 0);
+      if (confidence2 > 0) acc.confidences.push(confidence2);
+      return acc;
+    },
+    {
+      records: 0,
+      sources: 0,
+      documents: 0,
+      findings: 0,
+      duplicates: 0,
+      contradictions: 0,
+      lowTrustFiltered: 0,
+      confidences: [],
+      citationCoverage: [],
+      uniqueness: [],
+      relevance: [],
+      ragSuitability: [],
+      instructionTuning: [],
+      factualGrounding: [],
+      humanUsefulness: []
+    }
+  );
+  const confidence = totals.confidences.length ? totals.confidences.reduce((sum, value) => sum + value, 0) / totals.confidences.length : 0;
+  const citationCoverage = average(totals.citationCoverage);
+  const uniqueness = average(totals.uniqueness);
+  const relevance = average(totals.relevance);
+  const ragSuitability = average(totals.ragSuitability);
+  const instructionTuning = average(totals.instructionTuning);
+  const factualGrounding = average(totals.factualGrounding);
+  const humanUsefulness = average(totals.humanUsefulness);
+  console.log("");
+  console.log(paint("Alys run complete", "green"));
+  printStage("SRC", "DONE", "Authoritative sources ranked", formatInt(totals.sources));
+  printStage("SRC", "DONE", "Low-trust source filter applied", `${Math.round(totals.lowTrustFiltered / Math.max(1, response.datasets.length) * 100)}% avg filtered`);
+  printStage("EXT", "DONE", "Source documents normalized", formatInt(totals.documents));
+  printStage("CHK", "DONE", "Findings verified", formatInt(totals.findings));
+  printStage("CHK", "DONE", "Contradictory claims resolved", formatInt(totals.contradictions));
+  printStage("DED", "DONE", "Duplicate candidates removed", formatInt(totals.duplicates));
+  printStage("GEN", "DONE", "Canonical records accepted", formatInt(totals.records));
+  printStage("EVAL", "DONE", "Average confidence", formatPercent(confidence));
+  printStage("EVAL", "DONE", "Citation coverage", formatPercent(citationCoverage));
+  printStage("EVAL", "DONE", "Record uniqueness", formatPercent(uniqueness));
+  printStage("EVAL", "DONE", "Topic relevance", formatPercent(relevance));
+  printStage("EVAL", "DONE", "RAG suitability", formatScore(ragSuitability));
+  printStage("EVAL", "DONE", "Instruction tuning suitability", formatScore(instructionTuning));
+  printStage("EVAL", "DONE", "Factual grounding", formatScore(factualGrounding));
+  printStage("EVAL", "DONE", "Human usefulness proxy", formatScore(humanUsefulness));
+  printStage("OUT", "DONE", "Exports written", root);
+  console.log("");
+  console.log(paint("Datasets", "white"));
+  for (const dataset of response.datasets) {
+    const metrics = getMetrics(dataset);
+    const summary = getSummary(dataset);
+    const quality = getQualityMetrics(dataset);
+    const suitability = getEvaluation(dataset).suitability ?? getEvaluation(dataset).summary?.scores ?? {};
+    const records = Number(metrics.recordsGenerated ?? summary.recordsAccepted ?? 0);
+    const sources = Number(metrics.sourcesDiscovered ?? 0);
+    const confidenceValue = Number(metrics.averageConfidence ?? summary.averageConfidence ?? 0);
+    const outputDir = import_node_path.default.join(root, "datasets", dataset.id);
+    console.log(`${paint("\u2022", "yellow")} ${paint(dataset.id, "white")}  ${formatInt(records)} records  ${formatInt(sources)} sources  ${formatPercent(confidenceValue)} confidence`);
+    console.log(`  ${truncate(dataset.topic, 110)}`);
+    console.log(`  ${paint(outputDir, "cyan")}`);
+    console.log(`  quality ${formatPercent(Number(quality.citationCoverage ?? 0))} citations \xB7 ${formatPercent(Number(quality.recordUniqueness ?? 0))} unique \xB7 ${formatPercent(Number(quality.sourceDiversity ?? 0))} source diversity`);
+    console.log(`  suitability RAG ${formatScore(Number(suitability.ragSuitability ?? 0))} \xB7 tuning ${formatScore(Number(suitability.instructionTuning ?? 0))} \xB7 usefulness ${formatScore(Number(suitability.humanUsefulness ?? 0))}`);
+    const preview = previewRecord(dataset);
+    if (preview) {
+      console.log(paint("  preview", "gray"));
+      if (preview.input) console.log(`    in  ${paint(preview.input, "gray")}`);
+      if (preview.output) console.log(`    out ${preview.output}`);
+      for (const reason of preview.why) {
+        console.log(`    why ${paint(reason, "gray")}`);
+      }
+    }
+  }
+  console.log("");
+  console.log(
+    response.usage.isUnlimited ? paint("Generations remaining: unlimited", "green") : paint(`Generations remaining: ${response.usage.remaining ?? 0}`, "yellow")
+  );
+  console.log(paint(`Dashboard: ${appUrl()}/dashboard`, "cyan"));
+}
 async function handleStatus() {
   printBanner();
   const { profile } = await ensureAuthenticated();
@@ -5171,6 +5374,9 @@ async function handleLogout() {
 }
 async function handleGenerate(args, command) {
   const { config, profile } = await ensureAuthenticated();
+  printBanner();
+  printUsage(profile);
+  console.log("");
   const { values, positionals } = (0, import_node_util.parseArgs)({
     args: command === "generate" ? args.slice(1) : args,
     allowPositionals: true,
@@ -5187,13 +5393,13 @@ async function handleGenerate(args, command) {
       yes: { type: "boolean" }
     }
   });
-  const topicFromArgs = positionals.join(" ").trim();
-  const topic = topicFromArgs ? topicFromArgs : (await (0, import_prompts.default)({
+  const topicFromArgs = normalizeTopicInput(positionals.join(" "));
+  const topic = topicFromArgs ? topicFromArgs : normalizeTopicInput((await (0, import_prompts.default)({
     type: "text",
     name: "topic",
     message: "What do you want to generate?",
     validate: (v) => v.trim().length ? true : "Please enter a topic."
-  })).topic;
+  })).topic);
   if (!topic) throw new Error("Missing topic.");
   const datasetType = parseDatasetType(values.type) ?? (await (0, import_prompts.default)({
     type: "select",
@@ -5241,19 +5447,21 @@ async function handleGenerate(args, command) {
       { title: "Deep", value: "deep" }
     ]
   })).depth;
-  const sourceLimit = values.sources ? Math.max(1, Number(values.sources)) : (await (0, import_prompts.default)({
+  const sourceLimit = values.sources ? Math.min(MAX_SOURCES_PER_RUN, Math.max(1, Number(values.sources))) : (await (0, import_prompts.default)({
     type: "number",
     name: "sourceLimit",
     message: "How many sources?",
-    initial: 12,
-    min: 1
+    initial: MAX_SOURCES_PER_RUN,
+    min: 1,
+    max: MAX_SOURCES_PER_RUN
   })).sourceLimit;
-  const targetRows = values.rows ? Math.max(1, Number(values.rows)) : (await (0, import_prompts.default)({
+  const targetRows = values.rows ? Math.min(MAX_ROWS_PER_DATASET, Math.max(1, Number(values.rows))) : (await (0, import_prompts.default)({
     type: "number",
     name: "targetRows",
     message: "Rows per dataset?",
-    initial: 1e3,
-    min: 1
+    initial: MAX_ROWS_PER_DATASET,
+    min: 1,
+    max: MAX_ROWS_PER_DATASET
   })).targetRows;
   const workspaceRoot = (values.workspace ? String(values.workspace) : "").trim() || (await (0, import_prompts.default)({
     type: "text",
@@ -5265,13 +5473,11 @@ async function handleGenerate(args, command) {
   const verificationEnabled = values.verify === true ? true : values["no-verify"] === true ? false : (await (0, import_prompts.default)({
     type: "toggle",
     name: "verificationEnabled",
-    message: "Enable verification + debate swarm?",
+    message: "Enable verification checks?",
     initial: true,
     active: "Yes",
     inactive: "No"
   })).verificationEnabled;
-  printBanner();
-  printUsage(profile);
   printRunPlan({
     topic,
     datasetType,
@@ -5292,39 +5498,32 @@ async function handleGenerate(args, command) {
     console.log(paint("Run cancelled. No generations spent.", "green"));
     return;
   }
-  console.log(`${paint("[AUTH]", "gray")} ${paint("OK  ", "green")} Usage linked to ${appUrl()}`);
-  console.log(`${paint("[API ]", "gray")} ${paint("RUN ", "yellow")} Generating datasets through Akusa API...`);
-  const response = await requestJson(
-    "/api/cli/generate",
-    {
-      method: "POST",
-      body: JSON.stringify({
-        topic,
-        datasetType,
-        datasetCount,
-        formats: exportFormats,
-        sourceLimit,
-        targetRows,
-        depth,
-        verificationEnabled
-      })
-    },
-    config.token
-  );
-  await writeGeneratedDatasets(workspaceRoot, response.datasets);
-  console.log(`${paint("[API ]", "gray")} ${paint("DONE", "green")} Generation complete`);
-  console.log(`${paint("[OUT ]", "gray")} ${paint("DONE", "green")} Files written to ${resolveHome(workspaceRoot)}`);
-  console.log(paint("\nDatasets generated", "green"));
-  for (const dataset of response.datasets) {
-    const records = Number(dataset.manifest?.metrics && typeof dataset.manifest.metrics === "object" ? dataset.manifest.metrics.recordsGenerated ?? 0 : 0);
-    console.log(`${paint("\u2022", "yellow")} ${dataset.id}  records: ${formatInt(records)}`);
-    console.log(`  ${import_node_path.default.join(resolveHome(workspaceRoot), "datasets", dataset.id)}`);
-  }
-  console.log("");
-  console.log(
-    response.usage.isUnlimited ? paint("Generations remaining: unlimited", "green") : paint(`Generations remaining: ${response.usage.remaining ?? 0}`, "yellow")
+  console.log(paint("Runtime", "white"));
+  printStage("AUTH", "OK", "Usage linked", appUrl());
+  printStage("PLAN", "OK", "Generations charged only after successful completion", `${datasetCount} requested`);
+  printStage("SRC", "RUN", "Research pipeline starting", `${sourceLimit} source target`);
+  const response = await withSpinner(
+    "Alys research runtime executing",
+    requestJson(
+      "/api/cli/generate",
+      {
+        method: "POST",
+        body: JSON.stringify({
+          topic,
+          datasetType,
+          datasetCount,
+          formats: exportFormats,
+          sourceLimit,
+          targetRows,
+          depth,
+          verificationEnabled
+        })
+      },
+      config.token
+    )
   );
-  console.log(paint(`Dashboard: ${appUrl()}/dashboard`, "cyan"));
+  await withSpinner("Writing local exports", writeGeneratedDatasets(workspaceRoot, response.datasets));
+  printGenerationSummary(response, workspaceRoot);
 }
 function printFailure(error) {
   const message = error instanceof Error ? error.message : "Alys failed.";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "alys-akusa",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "private": false,
   "description": "Alys local CLI runtime for autonomous dataset generation.",
   "license": "UNLICENSED",