@appkit/llamacpp-cli 1.11.0 → 1.12.1

package/src/utils/process-utils.ts

@@ -1,316 +0,0 @@
-import { exec, spawn } from 'child_process';
-import { promisify } from 'util';
-
-export const execAsync = promisify(exec);
-
-/**
- * Execute a command and return stdout
- * Throws on non-zero exit code
- */
-export async function execCommand(command: string): Promise<string> {
-  const { stdout } = await execAsync(command);
-  return stdout.trim();
-}
-
-/**
- * Execute a command and return both stdout and stderr
- */
-export async function execCommandFull(command: string): Promise<{ stdout: string; stderr: string }> {
-  const { stdout, stderr } = await execAsync(command);
-  return {
-    stdout: stdout.trim(),
-    stderr: stderr.trim(),
-  };
-}
-
-/**
- * Check if a command exists in PATH
- */
-export async function commandExists(command: string): Promise<boolean> {
-  try {
-    await execAsync(`which ${command}`);
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/**
- * Check if a process is running by PID
- */
-export async function isProcessRunning(pid: number): Promise<boolean> {
-  try {
-    await execAsync(`ps -p ${pid}`);
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/**
- * Check if a port is in use
- */
-export async function isPortInUse(port: number): Promise<boolean> {
-  try {
-    await execAsync(`lsof -iTCP:${port} -sTCP:LISTEN -t`);
-    return true;
-  } catch {
-    return false;
-  }
-}
-
-/**
- * Spawn a streaming command, read one line, and kill it
- * Useful for commands like 'macmon pipe' that stream indefinitely
- * Ensures the process is killed to prevent leaks
- */
-export async function spawnAndReadOneLine(
-  command: string,
-  args: string[],
-  timeoutMs: number = 2000
-): Promise<string | null> {
-  return new Promise((resolve) => {
-    const child = spawn(command, args, {
-      stdio: ['ignore', 'pipe', 'ignore'],
-      detached: false, // Keep in same process group for easier cleanup
-    });
-
-    let resolved = false;
-    let output = '';
-
-    const cleanup = () => {
-      try {
-        // Try SIGKILL immediately (SIGTERM may not work for macmon)
-        child.kill('SIGKILL');
-      } catch {
-        // Process might already be dead
-      }
-    };
-
-    // Set timeout to kill process if it doesn't produce output
-    const timeout = setTimeout(() => {
-      if (!resolved) {
-        resolved = true;
-        cleanup();
-        resolve(null);
-      }
-    }, timeoutMs);
-
-    // Read stdout line by line
-    child.stdout?.on('data', (data) => {
-      if (resolved) return;
-
-      output += data.toString();
-
-      // Check if we have a complete line
-      const newlineIndex = output.indexOf('\n');
-      if (newlineIndex !== -1) {
-        const line = output.substring(0, newlineIndex).trim();
-
-        if (line.length > 0) {
-          resolved = true;
-          clearTimeout(timeout);
-          cleanup();
-          resolve(line);
-        }
-      }
-    });
-
-    // Handle process errors
-    child.on('error', () => {
-      if (!resolved) {
-        resolved = true;
-        clearTimeout(timeout);
-        resolve(null);
-      }
-    });
-
-    // Handle process exit
-    child.on('exit', () => {
-      if (!resolved) {
-        resolved = true;
-        clearTimeout(timeout);
-
-        // Return partial output if we have any
-        const line = output.trim();
-        resolve(line.length > 0 ? line : null);
-      }
-    });
-  });
-}
-
-// Process memory cache to prevent spawning too many 'top' processes
-// Cache per PID with 3-second TTL
-const processMemoryCache = new Map<number, { value: number | null; timestamp: number }>();
-const PROCESS_MEMORY_CACHE_TTL = 3000; // 3 seconds
-
-/**
- * Batch get memory usage for multiple processes in one top call
- * Much more efficient than calling getProcessMemory() multiple times
- * Returns Map<pid, bytes> for all requested PIDs
- */
-export async function getBatchProcessMemory(pids: number[]): Promise<Map<number, number | null>> {
-  const result = new Map<number, number | null>();
-  const now = Date.now();
-
-  // Check cache and collect PIDs that need fetching
-  const pidsToFetch: number[] = [];
-  for (const pid of pids) {
-    const cached = processMemoryCache.get(pid);
-    if (cached && (now - cached.timestamp) < PROCESS_MEMORY_CACHE_TTL) {
-      result.set(pid, cached.value);
-    } else {
-      pidsToFetch.push(pid);
-    }
-  }
-
-  // If all PIDs were cached, return early
-  if (pidsToFetch.length === 0) {
-    return result;
-  }
-
-  try {
-    // Build top command with all PIDs: top -l 1 -pid X -pid Y -pid Z -stats pid,mem
-    const pidArgs = pidsToFetch.map(pid => `-pid ${pid}`).join(' ');
-    const output = await execCommand(`top -l 1 ${pidArgs} -stats pid,mem 2>/dev/null`);
-
-    // Parse output: each line is "PID  MEM" (e.g., "1438  299M")
-    const lines = output.split('\n');
-    for (const line of lines) {
-      const match = line.trim().match(/^(\d+)\s+([\d.]+)([KMGT])\s*$/);
-      if (!match) continue;
-
-      const pid = parseInt(match[1], 10);
-      const value = parseFloat(match[2]);
-      const unit = match[3];
-
-      // Convert to bytes
-      const multipliers: { [key: string]: number } = {
-        K: 1024,
-        M: 1024 * 1024,
-        G: 1024 * 1024 * 1024,
-        T: 1024 * 1024 * 1024 * 1024,
-      };
-
-      const bytes = Math.round(value * multipliers[unit]);
-
-      // Cache and store result
-      processMemoryCache.set(pid, { value: bytes, timestamp: now });
-      result.set(pid, bytes);
-    }
-
-    // For any PIDs that weren't in the output, cache null
-    for (const pid of pidsToFetch) {
-      if (!result.has(pid)) {
-        processMemoryCache.set(pid, { value: null, timestamp: now });
-        result.set(pid, null);
-      }
-    }
-
-    return result;
-  } catch {
-    // On error, cache null for all requested PIDs
-    for (const pid of pidsToFetch) {
-      processMemoryCache.set(pid, { value: null, timestamp: now });
-      result.set(pid, null);
-    }
-    return result;
-  }
-}
-
-/**
- * Get memory usage for a single process in bytes
- * Uses 'top' on macOS which reports CPU memory only (NOT GPU/Metal memory)
- * Returns null if process not found or error occurs
- * Caches results for 3 seconds to prevent spawning too many top processes
- *
- * Note: For llama-server processes with GPU offloading, use ServerConfig.metalMemoryMB
- * to get GPU memory allocation (parsed from logs during server startup)
- *
- * Note: For multiple PIDs, use getBatchProcessMemory() instead - much more efficient
- */
-export async function getProcessMemory(pid: number): Promise<number | null> {
-  const result = await getBatchProcessMemory([pid]);
-  return result.get(pid) ?? null;
-}
-
-// Process CPU cache to prevent spawning too many 'ps' processes
-// Cache per PID with 3-second TTL
-const processCpuCache = new Map<number, { value: number | null; timestamp: number }>();
-const PROCESS_CPU_CACHE_TTL = 3000; // 3 seconds
-
-/**
- * Batch get CPU usage for multiple processes in one ps call
- * Much more efficient than calling getProcessCpu() multiple times
- * Returns Map<pid, percentage> for all requested PIDs
- */
-export async function getBatchProcessCpu(pids: number[]): Promise<Map<number, number | null>> {
-  const result = new Map<number, number | null>();
-  const now = Date.now();
-
-  // Check cache and collect PIDs that need fetching
-  const pidsToFetch: number[] = [];
-  for (const pid of pids) {
-    const cached = processCpuCache.get(pid);
-    if (cached && (now - cached.timestamp) < PROCESS_CPU_CACHE_TTL) {
-      result.set(pid, cached.value);
-    } else {
-      pidsToFetch.push(pid);
-    }
-  }
-
-  // If all PIDs were cached, return early
-  if (pidsToFetch.length === 0) {
-    return result;
-  }
-
-  try {
-    // Build ps command with all PIDs: ps -p X,Y,Z -o pid=,%cpu=
-    const pidList = pidsToFetch.join(',');
-    const output = await execCommand(`ps -p ${pidList} -o pid=,%cpu= 2>/dev/null`);
-
-    // Parse output: each line is "PID  %CPU" (e.g., "1438  45.2")
-    const lines = output.split('\n');
-    for (const line of lines) {
-      const match = line.trim().match(/^(\d+)\s+([\d.]+)\s*$/);
-      if (!match) continue;
-
-      const pid = parseInt(match[1], 10);
-      const cpuPercent = parseFloat(match[2]);
-
-      // Cache and store result
-      processCpuCache.set(pid, { value: cpuPercent, timestamp: now });
-      result.set(pid, cpuPercent);
-    }
-
-    // For any PIDs that weren't in the output, cache null (process not running)
-    for (const pid of pidsToFetch) {
-      if (!result.has(pid)) {
-        processCpuCache.set(pid, { value: null, timestamp: now });
-        result.set(pid, null);
-      }
-    }
-
-    return result;
-  } catch {
-    // On error, cache null for all requested PIDs
-    for (const pid of pidsToFetch) {
-      processCpuCache.set(pid, { value: null, timestamp: now });
-      result.set(pid, null);
-    }
-    return result;
-  }
-}
-
-/**
- * Get CPU usage for a single process as percentage (0-100+)
- * Uses 'ps -o %cpu' on macOS
- * Returns null if process not found or error occurs
- * Caches results for 3 seconds to prevent spawning too many ps processes
- *
- * Note: For multiple PIDs, use getBatchProcessCpu() instead - much more efficient
- */
-export async function getProcessCpu(pid: number): Promise<number | null> {
-  const result = await getBatchProcessCpu([pid]);
-  return result.get(pid) ?? null;
-}

package/src/utils/prompt-utils.ts

@@ -1,47 +0,0 @@
-import * as readline from 'readline';
-
-/**
- * Prompt user for input
- */
-export function prompt(question: string, defaultValue?: string): Promise<string> {
-  const rl = readline.createInterface({
-    input: process.stdin,
-    output: process.stdout,
-  });
-
-  return new Promise((resolve) => {
-    const promptText = defaultValue
-      ? `${question} [${defaultValue}]: `
-      : `${question}: `;
-
-    rl.question(promptText, (answer) => {
-      rl.close();
-      resolve(answer.trim() || defaultValue || '');
-    });
-  });
-}
-
-/**
- * Prompt user for yes/no confirmation
- */
-export function confirm(question: string, defaultYes = true): Promise<boolean> {
-  const rl = readline.createInterface({
-    input: process.stdin,
-    output: process.stdout,
-  });
-
-  const suffix = defaultYes ? '[Y/n]' : '[y/N]';
-
-  return new Promise((resolve) => {
-    rl.question(`${question} ${suffix}: `, (answer) => {
-      rl.close();
-      const input = answer.trim().toLowerCase();
-
-      if (input === '') {
-        resolve(defaultYes);
-      } else {
-        resolve(input === 'y' || input === 'yes');
-      }
-    });
-  });
-}

package/test-load.sh

@@ -1,100 +0,0 @@
-#!/bin/bash
-
-# Test script for parallel chat requests to multiple llama servers
-# Usage: ./test-load.sh
-# Stop with CTRL-C
-
-set -e
-
-# Available ports
-PORTS=(9001 9002 9004 9005)
-
-# Test prompts with varying complexity
-PROMPTS=(
-  "Write a hello world program in Python"
-  "Explain quantum computing in simple terms"
-  "Write a flappy bird game in Python"
-  "What are the benefits of functional programming?"
-  "Create a REST API example using FastAPI"
-  "Explain the difference between processes and threads"
-  "Write a binary search algorithm in JavaScript"
-  "What is the difference between HTTP and HTTPS?"
-  "Create a simple todo list app in React"
-  "Explain Docker containers to a beginner"
-  "Write a quicksort implementation in C++"
-  "What are the SOLID principles?"
-  "Create a SQL query to find duplicate records"
-  "Explain async/await in JavaScript"
-  "Write a Fibonacci sequence generator in any language"
-)
-
-# Colors for output
-RED='\033[0;31m'
-GREEN='\033[0;32m'
-YELLOW='\033[0;33m'
-BLUE='\033[0;34m'
-MAGENTA='\033[0;35m'
-CYAN='\033[0;36m'
-NC='\033[0m' # No Color
-
-# Counter for requests
-REQUEST_COUNT=0
-
-# Function to run a single chat request
-run_chat() {
-  local port=$1
-  local prompt=$2
-  local request_id=$3
-
-  echo -e "${CYAN}[Request #${request_id}]${NC} ${YELLOW}Port ${port}:${NC} ${prompt}"
-
-  # Run the chat request (suppress output except errors)
-  if npm run dev -- server run -m "$prompt" "$port" > /dev/null 2>&1; then
-    echo -e "${CYAN}[Request #${request_id}]${NC} ${GREEN}✓ Completed${NC} (port ${port})"
-  else
-    echo -e "${CYAN}[Request #${request_id}]${NC} ${RED}✗ Failed${NC} (port ${port})"
-  fi
-}
-
-# Trap CTRL-C for clean exit
-trap 'echo -e "\n${YELLOW}Stopping test load script...${NC}"; echo -e "${GREEN}Total requests sent: ${REQUEST_COUNT}${NC}"; exit 0' INT
-
-echo -e "${MAGENTA}========================================${NC}"
-echo -e "${MAGENTA}  Llama Server Load Test${NC}"
-echo -e "${MAGENTA}========================================${NC}"
-echo -e "${BLUE}Ports: ${PORTS[*]}${NC}"
-echo -e "${BLUE}Parallel requests: 3${NC}"
-echo -e "${BLUE}Press CTRL-C to stop${NC}"
-echo -e "${MAGENTA}========================================${NC}\n"
-
-# Function to count running background jobs
-count_running_jobs() {
-  jobs -r | wc -l | tr -d ' '
-}
-
-# Main loop - maintain exactly 3 concurrent requests at all times
-# Start initial 3 requests
-for i in {1..3}; do
-  PORT=${PORTS[$RANDOM % ${#PORTS[@]}]}
-  PROMPT=${PROMPTS[$RANDOM % ${#PROMPTS[@]}]}
-  ((REQUEST_COUNT++))
-  run_chat "$PORT" "$PROMPT" "$REQUEST_COUNT" &
-done
-
-# Continuously monitor and start new requests as old ones complete
-while true; do
-  # Get count of running background jobs
-  RUNNING=$(count_running_jobs)
-
-  # Start new requests to maintain 3 concurrent
-  while [ "$RUNNING" -lt 3 ]; do
-    PORT=${PORTS[$RANDOM % ${#PORTS[@]}]}
-    PROMPT=${PROMPTS[$RANDOM % ${#PROMPTS[@]}]}
-    ((REQUEST_COUNT++))
-    run_chat "$PORT" "$PROMPT" "$REQUEST_COUNT" &
-    RUNNING=$(count_running_jobs)
-  done
-
-  # Small sleep to avoid busy-waiting
-  sleep 0.5
-done

package/tsconfig.json

@@ -1,20 +0,0 @@
-{
-  "compilerOptions": {
-    "target": "ES2020",
-    "module": "commonjs",
-    "lib": ["ES2020"],
-    "outDir": "./dist",
-    "rootDir": "./src",
-    "strict": true,
-    "esModuleInterop": true,
-    "skipLibCheck": true,
-    "forceConsistentCasingInFileNames": true,
-    "resolveJsonModule": true,
-    "declaration": true,
-    "declarationMap": true,
-    "sourceMap": true,
-    "moduleResolution": "node"
-  },
-  "include": ["src/**/*"],
-  "exclude": ["node_modules", "dist"]
-}