@appkit/llamacpp-cli 1.8.0 → 1.10.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@appkit/llamacpp-cli",
-  "version": "1.8.0",
+  "version": "1.10.0",
   "description": "CLI tool to manage local llama.cpp servers on macOS",
   "main": "dist/cli.js",
   "bin": {

package/src/cli.ts

@@ -19,6 +19,12 @@ import { serverShowCommand } from './commands/server-show';
 import { serverConfigCommand } from './commands/config';
 import { configGlobalCommand } from './commands/config-global';
 import { monitorCommand } from './commands/monitor';
+import { routerStartCommand } from './commands/router/start';
+import { routerStopCommand } from './commands/router/stop';
+import { routerStatusCommand } from './commands/router/status';
+import { routerRestartCommand } from './commands/router/restart';
+import { routerConfigCommand } from './commands/router/config';
+import { routerLogsCommand } from './commands/router/logs';
 import packageJson from '../package.json';
 
 const program = new Command();
@@ -26,7 +32,17 @@ const program = new Command();
 program
   .name('llamacpp')
   .description('CLI tool to manage local llama.cpp servers on macOS')
-  .version(packageJson.version, '-v, --version', 'Output the version number');
+  .version(packageJson.version, '-v, --version', 'Output the version number')
+  .action(async () => {
+    // Default action: launch TUI when no command provided
+    try {
+      const { tuiCommand } = await import('./commands/tui');
+      await tuiCommand();
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
 
 // List models
 program
@@ -41,14 +57,13 @@ program
     }
   });
 
-// List running servers
+// List servers (static table)
 program
-  .command('ps [identifier]')
-  .description('Interactive server monitoring dashboard')
-  .option('--table', 'Show static table instead of TUI (for scripting)')
-  .action(async (identifier?: string, options?: { table?: boolean }) => {
+  .command('ps')
+  .description('List all servers with status (static table)')
+  .action(async () => {
     try {
-      await psCommand(identifier, options);
+      await psCommand();
     } catch (error) {
       console.error(chalk.red('❌ Error:'), (error as Error).message);
       process.exit(1);
@@ -293,14 +308,14 @@ server
     }
   });
 
-// Monitor server (deprecated - redirects to ps)
+// Monitor server (deprecated - redirects to TUI)
 server
   .command('monitor [identifier]')
-  .description('Monitor server with real-time metrics TUI (deprecated: use "llamacpp ps" instead)')
+  .description('Monitor server with real-time metrics TUI (deprecated: use "llamacpp" instead)')
   .action(async (identifier?: string) => {
     try {
       console.log(chalk.yellow('⚠️  The "monitor" command is deprecated and will be removed in a future version.'));
-      console.log(chalk.dim('   Please use "llamacpp ps" instead for the same functionality.\n'));
+      console.log(chalk.dim('   Please use "llamacpp" instead for the same functionality.\n'));
       await monitorCommand(identifier);
     } catch (error) {
       console.error(chalk.red('❌ Error:'), (error as Error).message);
@@ -308,5 +323,101 @@ server
     }
   });
 
+// Router management commands
+const router = program
+  .command('router')
+  .description('Manage the unified router endpoint');
+
+// Start router
+router
+  .command('start')
+  .description('Start the router service')
+  .action(async () => {
+    try {
+      await routerStartCommand();
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
+// Stop router
+router
+  .command('stop')
+  .description('Stop the router service')
+  .action(async () => {
+    try {
+      await routerStopCommand();
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
+// Show router status
+router
+  .command('status')
+  .description('Show router status and configuration')
+  .action(async () => {
+    try {
+      await routerStatusCommand();
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
+// Restart router
+router
+  .command('restart')
+  .description('Restart the router service')
+  .action(async () => {
+    try {
+      await routerRestartCommand();
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
+// Configure router
+router
+  .command('config')
+  .description('Update router configuration')
+  .option('-p, --port <number>', 'Update port number', parseInt)
+  .option('-h, --host <address>', 'Update bind address')
+  .option('--timeout <ms>', 'Update request timeout (milliseconds)', parseInt)
+  .option('--health-interval <ms>', 'Update health check interval (milliseconds)', parseInt)
+  .option('-v, --verbose [boolean]', 'Enable/disable verbose logging to file (true/false)', (val) => val === 'true' || val === '1')
+  .option('-r, --restart', 'Automatically restart router if running')
+  .action(async (options) => {
+    try {
+      await routerConfigCommand(options);
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
+// Router logs
+router
+  .command('logs')
+  .description('View router logs')
+  .option('-f, --follow', 'Follow logs in real-time (like tail -f)')
+  .option('-n, --lines <number>', 'Number of lines to show (default: 50)', parseInt)
+  .option('--stderr', 'Show system logs (stderr) instead of activity logs (stdout)')
+  .option('-v, --verbose', 'Show verbose JSON log file (if enabled)')
+  .option('--clear', 'Clear the log file')
+  .option('--rotate', 'Rotate the log file with timestamp')
+  .option('--clear-all', 'Clear all router logs (activity, system, verbose)')
+  .action(async (options) => {
+    try {
+      await routerLogsCommand(options);
+    } catch (error) {
+      console.error(chalk.red('❌ Error:'), (error as Error).message);
+      process.exit(1);
+    }
+  });
+
 // Parse arguments
 program.parse();

package/src/commands/monitor.ts

@@ -31,7 +31,7 @@ export async function monitorCommand(identifier?: string): Promise<void> {
       screen.destroy();
       throw new Error(
         `Server not found: ${identifier}\n\n` +
-          `Use: llamacpp ps\n` +
+          `Use: llamacpp ps (to list servers)\n` +
           `Or create a new server: llamacpp server create <model>`
       );
     }

package/src/commands/ps.ts

@@ -1,14 +1,32 @@
 import chalk from 'chalk';
 import Table from 'cli-table3';
-import blessed from 'blessed';
-import { stateManager } from '../lib/state-manager';
-import { statusChecker } from '../lib/status-checker';
-import { formatUptime, formatBytes } from '../utils/format-utils';
-import { getProcessMemory } from '../utils/process-utils';
-import { createMultiServerMonitorUI } from '../tui/MultiServerMonitorApp.js';
+import { stateManager } from '../lib/state-manager.js';
+import { statusChecker } from '../lib/status-checker.js';
+import { formatUptime, formatBytes } from '../utils/format-utils.js';
+import { getProcessMemory } from '../utils/process-utils.js';
 import { ServerConfig } from '../types/server-config.js';
 
-async function showStaticTable(): Promise<void> {
+const STATUS_CONFIG = {
+  running: { text: '✅ RUNNING', color: chalk.green },
+  crashed: { text: '❌ CRASHED', color: chalk.red },
+  stopped: { text: '⚠️  STOPPED', color: chalk.yellow },
+} as const;
+
+async function getServerMemory(server: ServerConfig): Promise<string> {
+  if (server.status !== 'running' || !server.pid) {
+    return '-';
+  }
+
+  const cpuMemoryBytes = await getProcessMemory(server.pid);
+  if (cpuMemoryBytes === null) {
+    return '-';
+  }
+
+  const metalMemoryBytes = server.metalMemoryMB ? server.metalMemoryMB * 1024 * 1024 : 0;
+  return formatBytes(cpuMemoryBytes + metalMemoryBytes);
+}
+
+export async function psCommand(): Promise<void> {
   const servers = await stateManager.getAllServers();
 
   if (servers.length === 0) {
@@ -17,60 +35,31 @@ async function showStaticTable(): Promise<void> {
     return;
   }
 
-  // Update all server statuses
   console.log(chalk.dim('Checking server statuses...\n'));
-  const updated = await statusChecker.updateAllServerStatuses();
+  const serversWithStatus = await statusChecker.updateAllServerStatuses();
 
   const table = new Table({
     head: ['SERVER ID', 'MODEL', 'PORT', 'STATUS', 'PID', 'MEMORY', 'UPTIME'],
   });
 
-  let runningCount = 0;
-  let stoppedCount = 0;
-  let crashedCount = 0;
-
-  for (const server of updated) {
-    let statusText: string;
-    let statusColor: (text: string) => string;
-
-    switch (server.status) {
-      case 'running':
-        statusText = '✅ RUNNING';
-        statusColor = chalk.green;
-        runningCount++;
-        break;
-      case 'crashed':
-        statusText = '❌ CRASHED';
-        statusColor = chalk.red;
-        crashedCount++;
-        break;
-      default:
-        statusText = '⚠️  STOPPED';
-        statusColor = chalk.yellow;
-        stoppedCount++;
-    }
-
-    const uptime =
-      server.status === 'running' && server.lastStarted
-        ? formatUptime(server.lastStarted)
-        : '-';
-
-    // Get memory usage for running servers (CPU + Metal GPU memory)
-    let memoryText = '-';
-    if (server.status === 'running' && server.pid) {
-      const cpuMemoryBytes = await getProcessMemory(server.pid);
-      if (cpuMemoryBytes !== null) {
-        const metalMemoryBytes = server.metalMemoryMB ? server.metalMemoryMB * 1024 * 1024 : 0;
-        const totalMemoryBytes = cpuMemoryBytes + metalMemoryBytes;
-        memoryText = formatBytes(totalMemoryBytes);
-      }
-    }
+  const counts = { running: 0, stopped: 0, crashed: 0 };
+
+  for (const server of serversWithStatus) {
+    const status = server.status || 'stopped';
+    const config = STATUS_CONFIG[status] || STATUS_CONFIG.stopped;
+    counts[status]++;
+
+    const uptime = server.status === 'running' && server.lastStarted
+      ? formatUptime(server.lastStarted)
+      : '-';
+
+    const memoryText = await getServerMemory(server);
 
     table.push([
       server.id,
       server.modelName,
       server.port.toString(),
-      statusColor(statusText),
+      config.color(config.text),
       server.pid?.toString() || '-',
       memoryText,
       uptime,
@@ -80,94 +69,16 @@ async function showStaticTable(): Promise<void> {
   console.log(table.toString());
 
   const summary = [
-    chalk.green(`${runningCount} running`),
-    chalk.yellow(`${stoppedCount} stopped`),
+    chalk.green(`${counts.running} running`),
+    chalk.yellow(`${counts.stopped} stopped`),
   ];
-  if (crashedCount > 0) {
-    summary.push(chalk.red(`${crashedCount} crashed`));
+  if (counts.crashed > 0) {
+    summary.push(chalk.red(`${counts.crashed} crashed`));
   }
 
   console.log(chalk.dim(`\nTotal: ${servers.length} servers (${summary.join(', ')})`));
 
-  if (crashedCount > 0) {
+  if (counts.crashed > 0) {
     console.log(chalk.red('\n⚠️  Some servers have crashed. Check logs with: llamacpp server logs <id> --errors'));
   }
 }
-
-export async function psCommand(identifier?: string, options?: { table?: boolean }): Promise<void> {
-  // If --table flag is set, show static table (backward compatibility)
-  if (options?.table) {
-    await showStaticTable();
-    return;
-  }
-
-  // Get all servers and update their statuses
-  const servers = await stateManager.getAllServers();
-
-  if (servers.length === 0) {
-    console.log(chalk.yellow('No servers configured.'));
-    console.log(chalk.dim('\nCreate a server: llamacpp server create <model-filename>'));
-    return;
-  }
-
-  // Update all server statuses
-  const updated = await statusChecker.updateAllServerStatuses();
-
-  // If identifier is provided, find the server and jump to detail view
-  if (identifier) {
-    const server = await findServer(identifier, updated);
-    if (!server) {
-      console.log(chalk.red(`❌ Server not found: ${identifier}`));
-      console.log(chalk.dim('\nAvailable servers:'));
-      updated.forEach((s: ServerConfig) => {
-        console.log(chalk.dim(`  - ${s.id} (port ${s.port})`));
-      });
-      process.exit(1);
-    }
-
-    // Find the server index for direct jump
-    const serverIndex = updated.findIndex(s => s.id === server.id);
-
-    // Launch multi-server TUI with direct jump to detail view
-    const screen = blessed.screen({
-      smartCSR: true,
-      title: 'llama.cpp Multi-Server Monitor',
-      fullUnicode: true,
-    });
-
-    await createMultiServerMonitorUI(screen, updated, true, serverIndex); // fromPs = true, directJumpIndex
-    return;
-  }
-
-  // No identifier - launch multi-server TUI
-  const runningServers = updated.filter((s: ServerConfig) => s.status === 'running');
-
-  // Launch multi-server TUI (shows all servers, not just running ones)
-  const screen = blessed.screen({
-    smartCSR: true,
-    title: 'llama.cpp Multi-Server Monitor',
-    fullUnicode: true,
-  });
-
-  await createMultiServerMonitorUI(screen, updated, true); // fromPs = true
-}
-
-// Helper function to find server by identifier
-async function findServer(identifier: string, servers: ServerConfig[]): Promise<ServerConfig | null> {
-  // Try by port
-  const port = parseInt(identifier);
-  if (!isNaN(port)) {
-    const server = servers.find(s => s.port === port);
-    if (server) return server;
-  }
-
-  // Try by exact ID
-  const byId = servers.find(s => s.id === identifier);
-  if (byId) return byId;
-
-  // Try by partial model name
-  const byModel = servers.find(s => s.modelName.toLowerCase().includes(identifier.toLowerCase()));
-  if (byModel) return byModel;
-
-  return null;
-}

package/src/commands/router/config.ts

@@ -0,0 +1,116 @@
+import chalk from 'chalk';
+import { routerManager } from '../../lib/router-manager';
+
+interface ConfigOptions {
+  port?: number;
+  host?: string;
+  timeout?: number;
+  healthInterval?: number;
+  verbose?: boolean;
+  restart?: boolean;
+}
+
+export async function routerConfigCommand(options: ConfigOptions): Promise<void> {
+  try {
+    // Check if router exists
+    const config = await routerManager.loadConfig();
+    if (!config) {
+      throw new Error('Router configuration not found. Use "llamacpp router start" to create it.');
+    }
+
+    // Check if any options were provided
+    const hasOptions = options.port || options.host || options.timeout || options.healthInterval || options.verbose !== undefined;
+    if (!hasOptions) {
+      throw new Error('No configuration options provided. Use --port, --host, --timeout, --health-interval, or --verbose');
+    }
+
+    const isRunning = config.status === 'running';
+
+    // Warn if running and no restart flag
+    if (isRunning && !options.restart) {
+      console.log(chalk.yellow('⚠️  Router is running. Changes will take effect after restart.'));
+      console.log(chalk.dim('   Use --restart flag to apply changes immediately.\n'));
+    }
+
+    // Prepare updates
+    const updates: any = {};
+    const changes: string[] = [];
+
+    if (options.port !== undefined) {
+      changes.push(`Port: ${config.port} → ${options.port}`);
+      updates.port = options.port;
+    }
+
+    if (options.host !== undefined) {
+      changes.push(`Host: ${config.host} → ${options.host}`);
+      updates.host = options.host;
+    }
+
+    if (options.timeout !== undefined) {
+      changes.push(`Request Timeout: ${config.requestTimeout}ms → ${options.timeout}ms`);
+      updates.requestTimeout = options.timeout;
+    }
+
+    if (options.healthInterval !== undefined) {
+      changes.push(`Health Check Interval: ${config.healthCheckInterval}ms → ${options.healthInterval}ms`);
+      updates.healthCheckInterval = options.healthInterval;
+    }
+
+    if (options.verbose !== undefined) {
+      const verboseStr = (val: boolean) => val ? 'enabled' : 'disabled';
+      changes.push(`Verbose Logging: ${verboseStr(config.verbose)} → ${verboseStr(options.verbose)}`);
+      updates.verbose = options.verbose;
+    }
+
+    // Display changes
+    console.log(chalk.blue('📝 Configuration changes:'));
+    console.log();
+    changes.forEach(change => {
+      console.log(chalk.dim(`  ${change}`));
+    });
+    console.log();
+
+    // Apply changes
+    if (isRunning && options.restart) {
+      console.log(chalk.blue('⏹️  Stopping router...'));
+      await routerManager.stop();
+    }
+
+    // Update config
+    await routerManager.updateConfig(updates);
+
+    // Regenerate plist if port or host changed
+    if (options.port !== undefined || options.host !== undefined) {
+      const updatedConfig = await routerManager.loadConfig();
+      if (updatedConfig) {
+        await routerManager.createPlist(updatedConfig);
+      }
+    }
+
+    // Restart if requested
+    if (isRunning && options.restart) {
+      console.log(chalk.blue('▶️  Starting router...'));
+      await routerManager.start();
+
+      const finalConfig = await routerManager.loadConfig();
+      console.log();
+      console.log(chalk.green('✅ Router restarted with new configuration'));
+      console.log();
+      console.log(chalk.dim(`Endpoint: http://${finalConfig?.host}:${finalConfig?.port}`));
+    } else {
+      console.log(chalk.green('✅ Configuration updated'));
+
+      if (isRunning) {
+        console.log();
+        console.log(chalk.yellow('⚠️  Restart required to apply changes:'));
+        console.log(chalk.dim('   llamacpp router restart'));
+      } else {
+        console.log();
+        console.log(chalk.dim('Start router to use new configuration:'));
+        console.log(chalk.dim('  llamacpp router start'));
+      }
+    }
+  } catch (error) {
+    throw new Error(`Failed to update router configuration: ${(error as Error).message}`);
+  }
+}