Eyal Toledano 21c3cb8cda feat(telemetry): Integrate telemetry for expand-all, aggregate results
This commit implements AI usage telemetry for the `expand-all-tasks` command/tool and refactors its CLI output for clarity and consistency.

Key Changes:

1.  **Telemetry Integration for `expand-all-tasks` (Subtask 77.8):**\n    -   The `expandAllTasks` core logic (`scripts/modules/task-manager/expand-all-tasks.js`) now calls the `expandTask` function for each eligible task and collects the individual `telemetryData` returned.\n    -   A new helper function `_aggregateTelemetry` (in `utils.js`) is used to sum up token counts and costs from all individual expansions into a single `telemetryData` object for the entire `expand-all` operation.\n    -   The `expandAllTasksDirect` wrapper (`mcp-server/src/core/direct-functions/expand-all-tasks.js`) now receives and passes this aggregated `telemetryData` in the MCP response.\n    -   For CLI usage, `displayAiUsageSummary` is called once with the aggregated telemetry.

2.  **Improved CLI Output for `expand-all`:**\n    -   The `expandAllTasks` core function now handles displaying a final "Expansion Summary" box (showing Attempted, Expanded, Skipped, Failed counts) directly after the aggregated telemetry summary.\n    -   This consolidates all summary output within the core function for better flow and removes redundant logging from the command action in `scripts/modules/commands.js`.\n    -   The summary box border is green for success and red if any expansions failed.

3.  **Code Refinements:**\n    -   Ensured `chalk` and `boxen` are imported in `expand-all-tasks.js` for the new summary box.\n    -   Minor adjustments to logging messages for clarity.
2025-05-08 18:22:00 -04:00

590 lines
17 KiB
JavaScript

/**
* utils.js
* Utility functions for the Task Master CLI
*/
import fs from 'fs';
import path from 'path';
import chalk from 'chalk';
import dotenv from 'dotenv';
// Import specific config getters needed here
import { getLogLevel, getDebugFlag } from './config-manager.js';
// Global silent mode flag
let silentMode = false;
// --- Environment Variable Resolution Utility ---
/**
* Resolves an environment variable's value.
* Precedence:
* 1. session.env (if session provided)
* 2. process.env
* 3. .env file at projectRoot (if projectRoot provided)
* @param {string} key - The environment variable key.
* @param {object|null} [session=null] - The MCP session object.
* @param {string|null} [projectRoot=null] - The project root directory (for .env fallback).
* @returns {string|undefined} The value of the environment variable or undefined if not found.
*/
function resolveEnvVariable(key, session = null, projectRoot = null) {
// 1. Check session.env
if (session?.env?.[key]) {
return session.env[key];
}
// 2. Read .env file at projectRoot
if (projectRoot) {
const envPath = path.join(projectRoot, '.env');
if (fs.existsSync(envPath)) {
try {
const envFileContent = fs.readFileSync(envPath, 'utf-8');
const parsedEnv = dotenv.parse(envFileContent); // Use dotenv to parse
if (parsedEnv && parsedEnv[key]) {
// console.log(`DEBUG: Found key ${key} in ${envPath}`); // Optional debug log
return parsedEnv[key];
}
} catch (error) {
// Log error but don't crash, just proceed as if key wasn't found in file
log('warn', `Could not read or parse ${envPath}: ${error.message}`);
}
}
}
// 3. Fallback: Check process.env
if (process.env[key]) {
return process.env[key];
}
// Not found anywhere
return undefined;
}
// --- Project Root Finding Utility ---
/**
* Finds the project root directory by searching upwards from a given starting point
* for a marker file or directory (e.g., 'package.json', '.git').
* @param {string} [startPath=process.cwd()] - The directory to start searching from.
* @param {string[]} [markers=['package.json', '.git', '.taskmasterconfig']] - Marker files/dirs to look for.
* @returns {string|null} The path to the project root directory, or null if not found.
*/
function findProjectRoot(
startPath = process.cwd(),
markers = ['package.json', '.git', '.taskmasterconfig']
) {
let currentPath = path.resolve(startPath);
while (true) {
for (const marker of markers) {
if (fs.existsSync(path.join(currentPath, marker))) {
return currentPath;
}
}
const parentPath = path.dirname(currentPath);
if (parentPath === currentPath) {
// Reached the filesystem root
return null;
}
currentPath = parentPath;
}
}
// --- Dynamic Configuration Function --- (REMOVED)
/*
function getConfig(session = null) {
// ... implementation removed ...
}
*/
// Set up logging based on log level
const LOG_LEVELS = {
debug: 0,
info: 1,
warn: 2,
error: 3,
success: 1 // Treat success like info level
};
/**
* Returns the task manager module
* @returns {Promise<Object>} The task manager module object
*/
async function getTaskManager() {
return import('./task-manager.js');
}
/**
* Enable silent logging mode
*/
function enableSilentMode() {
silentMode = true;
}
/**
* Disable silent logging mode
*/
function disableSilentMode() {
silentMode = false;
}
/**
* Check if silent mode is enabled
* @returns {boolean} True if silent mode is enabled
*/
function isSilentMode() {
return silentMode;
}
/**
* Logs a message at the specified level
* @param {string} level - The log level (debug, info, warn, error)
* @param {...any} args - Arguments to log
*/
function log(level, ...args) {
// Immediately return if silentMode is enabled
if (isSilentMode()) {
return;
}
// Get log level dynamically from config-manager
const configLevel = getLogLevel() || 'info'; // Use getter
// Use text prefixes instead of emojis
const prefixes = {
debug: chalk.gray('[DEBUG]'),
info: chalk.blue('[INFO]'),
warn: chalk.yellow('[WARN]'),
error: chalk.red('[ERROR]'),
success: chalk.green('[SUCCESS]')
};
// Ensure level exists, default to info if not
const currentLevel = LOG_LEVELS.hasOwnProperty(level) ? level : 'info';
// Check log level configuration
if (
LOG_LEVELS[currentLevel] >= (LOG_LEVELS[configLevel] ?? LOG_LEVELS.info)
) {
const prefix = prefixes[currentLevel] || '';
// Use console.log for all levels, let chalk handle coloring
// Construct the message properly
const message = args
.map((arg) => (typeof arg === 'object' ? JSON.stringify(arg) : arg))
.join(' ');
console.log(`${prefix} ${message}`);
}
}
/**
* Reads and parses a JSON file
* @param {string} filepath - Path to the JSON file
* @returns {Object|null} Parsed JSON data or null if error occurs
*/
function readJSON(filepath) {
// Get debug flag dynamically from config-manager
const isDebug = getDebugFlag();
try {
const rawData = fs.readFileSync(filepath, 'utf8');
return JSON.parse(rawData);
} catch (error) {
log('error', `Error reading JSON file ${filepath}:`, error.message);
if (isDebug) {
// Use dynamic debug flag
// Use log utility for debug output too
log('error', 'Full error details:', error);
}
return null;
}
}
/**
* Writes data to a JSON file
* @param {string} filepath - Path to the JSON file
* @param {Object} data - Data to write
*/
function writeJSON(filepath, data) {
// Get debug flag dynamically from config-manager
const isDebug = getDebugFlag();
try {
const dir = path.dirname(filepath);
if (!fs.existsSync(dir)) {
fs.mkdirSync(dir, { recursive: true });
}
fs.writeFileSync(filepath, JSON.stringify(data, null, 2), 'utf8');
} catch (error) {
log('error', `Error writing JSON file ${filepath}:`, error.message);
if (isDebug) {
// Use dynamic debug flag
// Use log utility for debug output too
log('error', 'Full error details:', error);
}
}
}
/**
* Sanitizes a prompt string for use in a shell command
* @param {string} prompt The prompt to sanitize
* @returns {string} Sanitized prompt
*/
function sanitizePrompt(prompt) {
// Replace double quotes with escaped double quotes
return prompt.replace(/"/g, '\\"');
}
/**
* Reads and parses the complexity report if it exists
* @param {string} customPath - Optional custom path to the report
* @returns {Object|null} The parsed complexity report or null if not found
*/
function readComplexityReport(customPath = null) {
// Get debug flag dynamically from config-manager
const isDebug = getDebugFlag();
try {
const reportPath =
customPath ||
path.join(process.cwd(), 'scripts', 'task-complexity-report.json');
if (!fs.existsSync(reportPath)) {
return null;
}
const reportData = fs.readFileSync(reportPath, 'utf8');
return JSON.parse(reportData);
} catch (error) {
log('warn', `Could not read complexity report: ${error.message}`);
// Optionally log full error in debug mode
if (isDebug) {
// Use dynamic debug flag
log('error', 'Full error details:', error);
}
return null;
}
}
/**
* Finds a task analysis in the complexity report
* @param {Object} report - The complexity report
* @param {number} taskId - The task ID to find
* @returns {Object|null} The task analysis or null if not found
*/
function findTaskInComplexityReport(report, taskId) {
if (
!report ||
!report.complexityAnalysis ||
!Array.isArray(report.complexityAnalysis)
) {
return null;
}
return report.complexityAnalysis.find((task) => task.taskId === taskId);
}
/**
* Checks if a task exists in the tasks array
* @param {Array} tasks - The tasks array
* @param {string|number} taskId - The task ID to check
* @returns {boolean} True if the task exists, false otherwise
*/
function taskExists(tasks, taskId) {
if (!taskId || !tasks || !Array.isArray(tasks)) {
return false;
}
// Handle both regular task IDs and subtask IDs (e.g., "1.2")
if (typeof taskId === 'string' && taskId.includes('.')) {
const [parentId, subtaskId] = taskId
.split('.')
.map((id) => parseInt(id, 10));
const parentTask = tasks.find((t) => t.id === parentId);
if (!parentTask || !parentTask.subtasks) {
return false;
}
return parentTask.subtasks.some((st) => st.id === subtaskId);
}
const id = parseInt(taskId, 10);
return tasks.some((t) => t.id === id);
}
/**
* Formats a task ID as a string
* @param {string|number} id - The task ID to format
* @returns {string} The formatted task ID
*/
function formatTaskId(id) {
if (typeof id === 'string' && id.includes('.')) {
return id; // Already formatted as a string with a dot (e.g., "1.2")
}
if (typeof id === 'number') {
return id.toString();
}
return id;
}
/**
* Finds a task by ID in the tasks array. Optionally filters subtasks by status.
* @param {Array} tasks - The tasks array
* @param {string|number} taskId - The task ID to find
* @param {string} [statusFilter] - Optional status to filter subtasks by
* @returns {{task: Object|null, originalSubtaskCount: number|null}} The task object (potentially with filtered subtasks) and the original subtask count if filtered, or nulls if not found.
*/
function findTaskById(tasks, taskId, statusFilter = null) {
if (!taskId || !tasks || !Array.isArray(tasks)) {
return { task: null, originalSubtaskCount: null };
}
// Check if it's a subtask ID (e.g., "1.2")
if (typeof taskId === 'string' && taskId.includes('.')) {
// If looking for a subtask, statusFilter doesn't apply directly here.
const [parentId, subtaskId] = taskId
.split('.')
.map((id) => parseInt(id, 10));
const parentTask = tasks.find((t) => t.id === parentId);
if (!parentTask || !parentTask.subtasks) {
return { task: null, originalSubtaskCount: null };
}
const subtask = parentTask.subtasks.find((st) => st.id === subtaskId);
if (subtask) {
// Add reference to parent task for context
subtask.parentTask = {
id: parentTask.id,
title: parentTask.title,
status: parentTask.status
};
subtask.isSubtask = true;
}
// Return the found subtask (or null) and null for originalSubtaskCount
return { task: subtask || null, originalSubtaskCount: null };
}
// Find the main task
const id = parseInt(taskId, 10);
const task = tasks.find((t) => t.id === id) || null;
// If task not found, return nulls
if (!task) {
return { task: null, originalSubtaskCount: null };
}
// If task found and statusFilter provided, filter its subtasks
if (statusFilter && task.subtasks && Array.isArray(task.subtasks)) {
const originalSubtaskCount = task.subtasks.length;
// Clone the task to avoid modifying the original array
const filteredTask = { ...task };
filteredTask.subtasks = task.subtasks.filter(
(subtask) =>
subtask.status &&
subtask.status.toLowerCase() === statusFilter.toLowerCase()
);
// Return the filtered task and the original count
return { task: filteredTask, originalSubtaskCount: originalSubtaskCount };
}
// Return original task and null count if no filter or no subtasks
return { task: task, originalSubtaskCount: null };
}
/**
* Truncates text to a specified length
* @param {string} text - The text to truncate
* @param {number} maxLength - The maximum length
* @returns {string} The truncated text
*/
function truncate(text, maxLength) {
if (!text || text.length <= maxLength) {
return text;
}
return text.slice(0, maxLength - 3) + '...';
}
/**
* Find cycles in a dependency graph using DFS
* @param {string} subtaskId - Current subtask ID
* @param {Map} dependencyMap - Map of subtask IDs to their dependencies
* @param {Set} visited - Set of visited nodes
* @param {Set} recursionStack - Set of nodes in current recursion stack
* @returns {Array} - List of dependency edges that need to be removed to break cycles
*/
function findCycles(
subtaskId,
dependencyMap,
visited = new Set(),
recursionStack = new Set(),
path = []
) {
// Mark the current node as visited and part of recursion stack
visited.add(subtaskId);
recursionStack.add(subtaskId);
path.push(subtaskId);
const cyclesToBreak = [];
// Get all dependencies of the current subtask
const dependencies = dependencyMap.get(subtaskId) || [];
// For each dependency
for (const depId of dependencies) {
// If not visited, recursively check for cycles
if (!visited.has(depId)) {
const cycles = findCycles(depId, dependencyMap, visited, recursionStack, [
...path
]);
cyclesToBreak.push(...cycles);
}
// If the dependency is in the recursion stack, we found a cycle
else if (recursionStack.has(depId)) {
// Find the position of the dependency in the path
const cycleStartIndex = path.indexOf(depId);
// The last edge in the cycle is what we want to remove
const cycleEdges = path.slice(cycleStartIndex);
// We'll remove the last edge in the cycle (the one that points back)
cyclesToBreak.push(depId);
}
}
// Remove the node from recursion stack before returning
recursionStack.delete(subtaskId);
return cyclesToBreak;
}
/**
* Convert a string from camelCase to kebab-case
* @param {string} str - The string to convert
* @returns {string} The kebab-case version of the string
*/
const toKebabCase = (str) => {
// Special handling for common acronyms
const withReplacedAcronyms = str
.replace(/ID/g, 'Id')
.replace(/API/g, 'Api')
.replace(/UI/g, 'Ui')
.replace(/URL/g, 'Url')
.replace(/URI/g, 'Uri')
.replace(/JSON/g, 'Json')
.replace(/XML/g, 'Xml')
.replace(/HTML/g, 'Html')
.replace(/CSS/g, 'Css');
// Insert hyphens before capital letters and convert to lowercase
return withReplacedAcronyms
.replace(/([A-Z])/g, '-$1')
.toLowerCase()
.replace(/^-/, ''); // Remove leading hyphen if present
};
/**
* Detect camelCase flags in command arguments
* @param {string[]} args - Command line arguments to check
* @returns {Array<{original: string, kebabCase: string}>} - List of flags that should be converted
*/
function detectCamelCaseFlags(args) {
const camelCaseFlags = [];
for (const arg of args) {
if (arg.startsWith('--')) {
const flagName = arg.split('=')[0].slice(2); // Remove -- and anything after =
// Skip single-word flags - they can't be camelCase
if (!flagName.includes('-') && !/[A-Z]/.test(flagName)) {
continue;
}
// Check for camelCase pattern (lowercase followed by uppercase)
if (/[a-z][A-Z]/.test(flagName)) {
const kebabVersion = toKebabCase(flagName);
if (kebabVersion !== flagName) {
camelCaseFlags.push({
original: flagName,
kebabCase: kebabVersion
});
}
}
}
}
return camelCaseFlags;
}
/**
* Aggregates an array of telemetry objects into a single summary object.
* @param {Array<Object>} telemetryArray - Array of telemetryData objects.
* @param {string} overallCommandName - The name for the aggregated command.
* @returns {Object|null} Aggregated telemetry object or null if input is empty.
*/
function _aggregateTelemetry(telemetryArray, overallCommandName) {
if (!telemetryArray || telemetryArray.length === 0) {
return null;
}
const aggregated = {
timestamp: new Date().toISOString(), // Use current time for aggregation time
userId: telemetryArray[0].userId, // Assume userId is consistent
commandName: overallCommandName,
modelUsed: 'Multiple', // Default if models vary
providerName: 'Multiple', // Default if providers vary
inputTokens: 0,
outputTokens: 0,
totalTokens: 0,
totalCost: 0,
currency: telemetryArray[0].currency || 'USD' // Assume consistent currency or default
};
const uniqueModels = new Set();
const uniqueProviders = new Set();
const uniqueCurrencies = new Set();
telemetryArray.forEach((item) => {
aggregated.inputTokens += item.inputTokens || 0;
aggregated.outputTokens += item.outputTokens || 0;
aggregated.totalCost += item.totalCost || 0;
uniqueModels.add(item.modelUsed);
uniqueProviders.add(item.providerName);
uniqueCurrencies.add(item.currency || 'USD');
});
aggregated.totalTokens = aggregated.inputTokens + aggregated.outputTokens;
aggregated.totalCost = parseFloat(aggregated.totalCost.toFixed(6)); // Fix precision
if (uniqueModels.size === 1) {
aggregated.modelUsed = [...uniqueModels][0];
}
if (uniqueProviders.size === 1) {
aggregated.providerName = [...uniqueProviders][0];
}
if (uniqueCurrencies.size > 1) {
aggregated.currency = 'Multiple'; // Mark if currencies actually differ
} else if (uniqueCurrencies.size === 1) {
aggregated.currency = [...uniqueCurrencies][0];
}
return aggregated;
}
// Export all utility functions and configuration
export {
LOG_LEVELS,
log,
readJSON,
writeJSON,
sanitizePrompt,
readComplexityReport,
findTaskInComplexityReport,
taskExists,
formatTaskId,
findTaskById,
truncate,
findCycles,
toKebabCase,
detectCamelCaseFlags,
enableSilentMode,
disableSilentMode,
isSilentMode,
resolveEnvVariable,
getTaskManager,
findProjectRoot,
_aggregateTelemetry
};