kiku-agent 1.3.2 → 1.3.4

package/server/agent/ui.js

@@ -645,6 +645,9 @@ export function formatStatusBadge(statusType) {
       return `${C.cyberCyan}${C.bold}[◆ INFO]${C.reset}`;
     case 'mcp':
       return `${C.lavender}${C.bold}[● MCP]${C.reset}`;
+    case 'session':
+    case 'session_restored':
+      return `\x1b[48;2;236;72;153m\x1b[38;2;24;6;16m${C.bold} ✦ SESSION RESTORED ${C.reset}`;
     default:
       return `${C.slate}${C.bold}[${String(statusType).toUpperCase()}]${C.reset}`;
   }
@@ -753,6 +756,18 @@ export const STATUS_THEMES = Object.freeze({
     bodyBg: '\x1b[48;2;10;32;46m', bodyFg: '\x1b[38;2;186;230;253m',
     icon: '◆', defaultBadge: 'INFO',
   },
+  session: {
+    badgeBg: '\x1b[48;2;236;72;153m', badgeFg: '\x1b[38;2;24;6;16m',
+    headerBg: '\x1b[48;2;64;18;42m', headerFg: '\x1b[38;2;253;242;248m',
+    bodyBg: '\x1b[48;2;42;12;28m', bodyFg: '\x1b[38;2;251;207;232m',
+    icon: '✦', defaultBadge: 'SESSION RESTORED',
+  },
+  session_restored: {
+    badgeBg: '\x1b[48;2;236;72;153m', badgeFg: '\x1b[38;2;24;6;16m',
+    headerBg: '\x1b[48;2;64;18;42m', headerFg: '\x1b[38;2;253;242;248m',
+    bodyBg: '\x1b[48;2;42;12;28m', bodyFg: '\x1b[38;2;251;207;232m',
+    icon: '✦', defaultBadge: 'SESSION RESTORED',
+  },
 });
 
 export function formatRichCardLines({
@@ -1122,61 +1137,51 @@ export function getWelcomeHeroLines() {
   const r = C.roseGlow;
   const g = C.gold;
   const c = C.cyberCyan;
-  const l = C.lavender;
-  const e = C.emerald;
   const s = C.slate;
   const w = C.white;
   const b = C.bold;
   const d = C.dim;
   const rst = C.reset;
-  const boxWidth = Math.min(84, Math.max(50, getFullWidth() - 4));
-  const compact = getFullWidth() < 120;
-  const headerTitle = 'Suggested Topics & Queries';
-  const topDashCount = Math.max(2, boxWidth - getStringDisplayWidth(headerTitle) - 6);
+  const fullWidth = getFullWidth();
+  const boxWidth = Math.min(80, Math.max(48, fullWidth - 4));
   const innerWidth = boxWidth - 2;
+  const headerTitle = 'Suggested Queries';
+  const topDashCount = Math.max(2, boxWidth - getStringDisplayWidth(headerTitle) - 6);
 
   const padItem = (icon, tag, text) => {
     const tagWidth = 14;
-    const tagPadded = tag + ' '.repeat(Math.max(0, tagWidth - getStringDisplayWidth(tag)));
     const visualContent = 2 + 2 + tagWidth + 2 + getStringDisplayWidth(text);
     const pad = Math.max(0, innerWidth - visualContent);
     return `  ${s}│${rst}  ${icon} ${tag}${rst}${' '.repeat(Math.max(0, tagWidth - getStringDisplayWidth(tag)))}  ${w}${text}${rst}${' '.repeat(pad)}${s}│${rst}`;
   };
 
-  const heroLines = compact ? [
-    `  ${p}██╗  ██╗██╗██╗  ██╗██╗   ██╗     █████╗ ██╗${rst}`,
-    `  ${r}█████═╝ ██║█████═╝ ██║   ██║    ███████║██║${rst}`,
-    `  ${p}██║  ██║██║██║  ██║╚██████╔╝    ██║  ██║██║${rst}`,
+  const heroLines = fullWidth < 78 ? [
+    `  ${p}██╗  ██╗██╗██╗  ██╗██╗   ██╗${rst}`,
+    `  ${r}█████═╝ ██║█████═╝ ██║   ██║${rst}`,
+    `  ${p}██║  ██║██║██║  ██║╚██████╔╝${rst}`,
+    `  ${r}◆ ${b}${w}KIKU AGENT${rst} ${s}•${rst} ${d}[Sakura]${rst}`,
   ] : [
-    `  ${p}██╗  ██╗██╗██╗  ██╗██╗   ██╗     █████╗ ██╗${rst}`,
-    `  ${p}██║ ██╔╝██║██║ ██╔╝██║   ██║    ██╔══██╗██║${rst}`,
-    `  ${r}█████═╝ ██║█████═╝ ██║   ██║    ███████║██║${rst}`,
-    `  ${r}██╔═██╗ ██║██╔═██╗ ██║   ██║    ██╔══██╗██║${rst}`,
-    `  ${p}██║ ╚██╗██║██║ ╚██╗╚██████╔╝    ██║  ██║██║${rst}`,
-    `  ${p}╚═╝  ╚═╝╚═╝╚═╝  ╚═╝ ╚═════╝     ╚═╝  ╚═╝╚═╝${rst}`,
-  ];
-  const shortcutLines = compact ? [
-    `  ${s}Keys: ${b}${w}[Tab]${rst} ${s}Cmds • ${b}${w}[↑/↓]${rst} ${s}Hist • ${b}${w}[F1]${rst} ${s}Help • ${b}${w}[F2]${rst} ${s}Details${rst}`,
-    `  ${b}${w}[F3]${rst} ${s}Think • ${b}${w}[F4]${rst} ${s}Tools • ${b}${w}[F5]${rst} ${s}Copy • ${b}${w}[Ctrl+C]${rst} ${s}Exit${rst}`,
-  ] : [
-    `  ${s}Keys:  ${b}${w}[Tab]${rst} ${s}Autocomplete • ${b}${w}[↑/↓]${rst} ${s}History • ${b}${w}[PgUp/PgDn]${rst} ${s}Scroll • ${b}${w}[Ctrl+C]${rst} ${s}Exit${rst}`,
-    `  ${s}F-Keys: ${b}${w}[F1]${rst} ${s}Help • ${b}${w}[F2]${rst} ${s}Details • ${b}${w}[F3]${rst} ${s}Thinking • ${b}${w}[F4]${rst} ${s}Tools • ${b}${w}[F5]${rst} ${s}Copy • ${b}${w}[F6]${rst} ${s}Gallery • ${b}${w}[F7]${rst} ${s}Theme${rst}`,
-    `  ${s}Ctrl:  ${b}${w}[Ctrl+T]${rst} ${s}Toggle Think • ${b}${w}[Ctrl+R]${rst} ${s}Retry • ${b}${w}[Ctrl+S]${rst} ${s}Save • ${b}${w}[Ctrl+G]${rst} ${s}Gallery${rst}`,
+    `  ${p}██╗  ██╗██╗██╗  ██╗██╗   ██╗    ${r}█████╗  ██████╗ ███████╗███╗   ██╗████████╗${rst}`,
+    `  ${p}██║ ██╔╝██║██║ ██╔╝██║   ██║   ${r}██╔══██╗██╔════╝ ██╔════╝████╗  ██║╚══██╔══╝${rst}`,
+    `  ${r}█████═╝ ██║█████═╝ ██║   ██║   ${r}███████║██║  ███╗█████╗  ██╔██╗ ██║   ██║   ${rst}`,
+    `  ${r}██╔═██╗ ██║██╔═██╗ ██║   ██║   ${p}██╔══██║██║   ██║██╔══╝  ██║╚██╗██║   ██║   ${rst}`,
+    `  ${p}██║ ╚██╗██║██║ ╚██╗╚██████╔╝   ${p}██║  ██║╚██████╔╝███████╗██║ ╚████║   ██║   ${rst}`,
   ];
 
+  const subtitle = `  ${p}🌸 ${b}${w}Ju Jingyi (鞠婧祎) AI Concierge${rst} ${s}•${rst} ${d}Interactive Terminal Agent${rst}`;
+
+  const shortcutLine = fullWidth < 85
+    ? `  ${s}Keys: ${b}${w}[Tab]${rst} ${s}Cmds • ${b}${w}[↑/↓]${rst} ${s}Hist • ${b}${w}[F1]${rst} ${s}Help • ${b}${w}[Ctrl+C]${rst} ${s}Exit${rst}`
+    : `  ${s}Keys: ${b}${w}[Tab]${rst} ${s}Autocomplete • ${b}${w}[↑/↓]${rst} ${s}History • ${b}${w}[F1]${rst} ${s}Help • ${b}${w}[F3]${rst} ${s}Think • ${b}${w}[Ctrl+C]${rst} ${s}Exit${rst}`;
+
   return [
-    ...(compact ? [] : ['']),
     ...heroLines,
-    '',
+    subtitle,
     `  ${s}╭──${rst} ${b}${g}${headerTitle}${rst} ${s}${'─'.repeat(topDashCount)}╮${rst}`,
-    padItem(p + '◆', '2026 Works', '"What are Ju Jingyi\'s 2026 upcoming dramas?"'),
-    padItem(c + '◆', 'Music & OST', '"OSTs from In Blossom & Ruyi Pavilion, /play Ginkgo"'),
-    padItem(e + '◆', 'Endorsements', '"Versace Global Ambassador & fashion partnerships"'),
-    padItem(l + '◆', 'Tools / MCP', '"@search_works 2026, /mcp list, /stats, /help"'),
+    padItem(p + '◆', 'Ju Jingyi Works', '"What are Ju Jingyi\'s 2026 upcoming dramas?"'),
+    padItem(c + '◆', 'Music & Tools', '"OSTs from In Blossom, /play Ginkgo, @search_works"'),
     `  ${s}╰${'─'.repeat(innerWidth)}╯${rst}`,
-    '',
-    ...shortcutLines,
-    '',
+    shortcutLine,
   ];
 }
 

package/server/agent/vramService.js

@@ -13,6 +13,8 @@ import { getGpuMetrics } from './gpuMetrics.js';
 import { listInstalledGgufs, resolveModelConfig } from './modelConfig.js';
 import { classifyHardwareTier, calculateDynamicVramBudget, HARDWARE_TIERS } from './telemetry/hardwareTiers.js';
 import { techPowerUpClient } from './telemetry/techPowerUpClient.js';
+import { probeLmStudioRuntime } from './lmStudioWakeup.js';
+import { getCpuMetrics } from './cpuMetrics.js';
 
 function runProcessCapture(cmd, args, timeoutMs = 2500) {
   return new Promise((resolve) => {
@@ -154,12 +156,43 @@ export async function getHardwareTarget() {
 
 
 export async function buildHardwareReport() {
-  const [target, probe] = await Promise.all([
+  const [target, probe, cpu] = await Promise.all([
     getHardwareTarget(),
     getGpuMetrics().catch(() => ({ available: false })),
+    getCpuMetrics().catch(() => null),
   ]);
+
+  const primaryGpu = probe.available && probe.gpus?.[0] ? probe.gpus[0] : null;
+  const gpuName = primaryGpu?.name || target.name || '';
+  const vramGb = primaryGpu?.totalMb ? primaryGpu.totalMb / 1024 : (target.totalBytes ? target.totalBytes / 1024 ** 3 : 0);
+
+  let gpuSpecs = null;
+  try {
+    gpuSpecs = techPowerUpClient.disambiguateGpu(gpuName, vramGb);
+  } catch {}
+
+  const gpu = primaryGpu ? {
+    name: gpuSpecs?.name || gpuName,
+    vendor: primaryGpu.vendor || target.vendor || 'Unknown',
+    vramTotalMb: primaryGpu.totalMb ?? null,
+    vramUsedMb: primaryGpu.usedMb ?? null,
+    vramFreeMb: primaryGpu.freeMb ?? null,
+    vramTotalGb: vramGb ? Number(vramGb.toFixed(1)) : null,
+    utilizationGpu: primaryGpu.utilizationGpu ?? null,
+    temperatureC: primaryGpu.temperatureC ?? null,
+    driverVersion: primaryGpu.driverVersion ?? null,
+    busWidth: gpuSpecs?.busBits ?? null,
+    bandwidthGbps: gpuSpecs?.bandwidthGbs ?? null,
+    memoryType: gpuSpecs?.memoryType ?? null,
+    architecture: gpuSpecs?.arch ?? null,
+    pcieGeneration: gpuSpecs?.pcie ?? null,
+    source: primaryGpu.source || target.source || 'gpu-probe',
+  } : null;
+
   return {
     available: Boolean(target.available),
+    gpu,
+    cpu,
     installedModels: listInstalledGgufs(),
     target: {
       vendor: target.vendor,
@@ -242,59 +275,7 @@ function pickFirst(...values) {
 
 
 
-const LMSTUDIO_PROBE_TTL_MS = 15_000;
-const lmStudioProbeCache = new Map();
-
-export async function probeLmStudioRuntime(baseUrl = 'http://127.0.0.1:1234/v1') {
-  const key = String(baseUrl || 'http://127.0.0.1:1234/v1');
-  const cached = lmStudioProbeCache.get(key);
-  if (cached) {
-    if (cached.inFlight) return cached.inFlight;
-    if (Date.now() - cached.fetchedAt < LMSTUDIO_PROBE_TTL_MS) return cached.data;
-  }
-  const entry = { fetchedAt: Date.now(), data: null, inFlight: probeLmStudioRuntimeUncached(baseUrl) };
-  lmStudioProbeCache.set(key, entry);
-  try {
-    const result = await entry.inFlight;
-    entry.data = result;
-    entry.fetchedAt = Date.now();
-    return result;
-  } catch (err) {
-    lmStudioProbeCache.delete(key);
-    throw err;
-  } finally {
-    entry.inFlight = null;
-  }
-}
-
-async function probeLmStudioRuntimeUncached(baseUrl = 'http://127.0.0.1:1234/v1') {
-  const result = { reachable: false, loadedModel: null, loadedModelType: null, loadedContextLength: null, quantization: null, pendingModel: null };
-  try {
-    const root = String(baseUrl || 'http://127.0.0.1:1234/v1').replace(/\/v1\/?$/, '');
-    const res = await fetch(`${root}/api/v0/models`, { signal: AbortSignal.timeout(2500) });
-    if (!res.ok) return result;
-    const data = await res.json();
-    const models = Array.isArray(data?.data) ? data.data : [];
-    if (models.length === 0) return result;
-    result.reachable = true;
-    const loaded = models.find(m => m.state === 'loaded' && (m.type === 'llm' || m.type === 'vlm'))
-      || models.find(m => m.state === 'loaded');
-    if (loaded) {
-      result.loadedModel = loaded.id || null;
-      result.loadedModelType = loaded.type || null;
-      result.loadedContextLength = loaded.loaded_context_length ?? loaded.max_context_length ?? null;
-      result.quantization = loaded.quantization || null;
-      return result;
-    }
-  
-    const pending = models.find(m => m.state === 'loading' && (m.type === 'llm' || m.type === 'vlm'))
-      || models.find(m => m.state === 'loading');
-    if (pending) result.pendingModel = pending.id || null;
-    return result;
-  } catch {
-    return result;
-  }
-}
+export { probeLmStudioRuntime } from './lmStudioWakeup.js';
 
 export async function computeVramEstimate(request = {}) {
   let modelName = request.modelName || request.model || request.model_id || null;