const fs = require('fs')
const path = require('path')
const glob = require('glob')

// Function to convert <a href="url">text</a> to [text](url)
function convertLinksInFile(filePath) {
  console.log(`Processing: ${filePath}`)
  let content = fs.readFileSync(filePath, 'utf8')

  // Count links before conversion
  const linkMatches = content.match(/<a href="[^"]*"[^>]*>[^<]*<\/a>/g) || []
  const linkCount = linkMatches.length

  // Replace <a href="url" target="_blank">text</a> with [text](url)
  const updatedContent = content.replace(
    /<a href="([^"]*)"(?:[^>]*?)>([^<]*)<\/a>/g,
    '[$2]($1)',
  )

  // Count links after conversion
  const markdownLinkMatches = updatedContent.match(/\[[^\]]*\]$[^)]*$/g) || []
  const markdownLinkCount = markdownLinkMatches.length

  // Write the updated content back to the file if changes were made
  if (content !== updatedContent) {
    fs.writeFileSync(filePath, updatedContent, 'utf8')
    console.log(`✅ Converted ${linkCount} links in ${filePath}`)
    console.log(`   Found ${markdownLinkCount} markdown links after conversion`)
    return {
      file: filePath,
      linksConverted: linkCount,
      markdownLinksAfter: markdownLinkCount,
    }
  } else {
    console.log(`ℹ️ No changes made to ${filePath}`)
    return {
      file: filePath,
      linksConverted: 0,
      markdownLinksAfter: markdownLinkCount,
    }
  }
}

// Process either a single file or all MDX files
const processFiles = (pattern) => {
  const files = glob.sync(pattern)
  console.log(`Found ${files.length} files matching pattern: ${pattern}`)

  let totalConverted = 0
  const summary = []

  files.forEach((file) => {
    const result = convertLinksInFile(file)
    totalConverted += result.linksConverted
    summary.push(result)
  })

  console.log('\n--- Summary ---')
  console.log(`Total links converted: ${totalConverted}`)

  // Files with most conversions (top 5)
  const sortedSummary = [...summary].sort(
    (a, b) => b.linksConverted - a.linksConverted,
  )
  console.log('\nTop files by number of conversions:')
  sortedSummary.slice(0, 5).forEach((item) => {
    if (item.linksConverted > 0) {
      console.log(`${item.file}: ${item.linksConverted} links converted`)
    }
  })

  // Create conversion log file
  const logContent = `# Link Conversion Log
  
Date: ${new Date().toISOString().split('T')[0]}

## Summary
- Total files processed: ${files.length}
- Total links converted: ${totalConverted}

## Files with most conversions
${sortedSummary
  .slice(0, 10)
  .map((item) =>
    item.linksConverted > 0
      ? `- ${item.file}: ${item.linksConverted} links converted`
      : '',
  )
  .filter((line) => line !== '')
  .join('\n')}

## All converted files
${summary
  .filter((item) => item.linksConverted > 0)
  .map((item) => `- ${item.file}: ${item.linksConverted} links converted`)
  .join('\n')}
`

  fs.writeFileSync(
    path.join(__dirname, 'conversion-log.md'),
    logContent,
    'utf8',
  )
  console.log('\nConversion log saved to utils/conversion-log.md')

  return { totalConverted, summary }
}

// Check arguments
const args = process.argv.slice(2)
if (args[0] === '--test') {
  // Test mode: process only one file
  if (args[1]) {
    console.log('TEST MODE: Processing single file')
    processFiles(args[1])
  } else {
    console.log(
      'Please provide a test file path, e.g.: node convert-links.js --test ./some-file.mdx',
    )
  }
} else if (args[0] === '--dry-run') {
  // Dry run mode: just report what would be changed
  console.log('DRY RUN MODE: Not making actual changes')
  const originalWriteFileSync = fs.writeFileSync
  fs.writeFileSync = (path, data, options) => {
    console.log(`Would write to ${path} (changes suppressed in dry run)`)
  }
  processFiles('./**/*.mdx')
  fs.writeFileSync = originalWriteFileSync
} else {
  // Process all MDX files in current directory
  processFiles('./**/*.mdx')
}


Quickstart

Steps to Detect Anomalies

Next Steps

Get started with TimeGPT's historical anomaly detection capabilities.

Historical Anomaly Detection Quickstart

TimeGPT Foundational model for time series forecasting and anomaly detection

Welcome to TimeGPT - The foundational model for time series forecasting and anomaly detection

Introduction

Learn how to use TimeGPT for accurate time series forecasting in just a few steps

Learn about TimeGPT - the foundation model for time series.

About TimeGPT

Understand the benefits of using TimeGPT for time series analysis.

Why TimeGPT?

TimeGPT FAQ

Get started with TimeGPT using Polars for efficient data processing.

TimeGPT Quickstart (Polars)

Quickstart guide to deploy and use TimeGEN-1 on Azure with the Nixtla Python SDK for time series forecasting.

TimeGEN-1 Quickstart (Azure)

Learn how to securely configure your Nixtla SDK API key using direct code or environment variables.

Setting up your API key

Overview of TimeGPT's Enterprise subscription plans with deployment options, support, and trial details.

TimeGPT Subscription Plans

Overview of the data format and requirements for TimeGPT forecasting.

Data Requirements

Key terminology and concepts for time series forecasting with TimeGPT

Glossary

Complete list of changes for each version of the Nixtla client.

Changelog

Advanced zero-shot forecasting capabilities for time series data

Forecast

Get started quickly with TimeGPT forecasting using the Nixtla API.

Learn how to incorporate external factors to improve forecast accuracy.

Add Exogenous Variables

Incorporate country-specific holidays and custom special dates into your forecasting models to improve accuracy.

Add holidays and special dates

Learn how to create and use custom categorical variables with SpecialDates in TimeGPT forecasting.

Add categorical variables

Learn how to predict multiple seasonal cycles into the future using TimeGPT

Long-horizon forecasting

Learn how to forecast multiple time series at once with TimeGPT.

Multiple Series Forecasting

How to fine-tune TimeGPT forecasts using finetune_steps and finetune_depth parameters for improved accuracy.

Fine-tuning

Learn how to fine-tune forecasting models using a preferred loss function for improved performance.

Finetuning with a custom loss function

Learn how to validate your forecast models using time series cross-validation.

Cross Validation

Learn how to use the level parameter to generate prediction intervals that quantify forecast uncertainty.

Prediction Intervals

Learn how to work with both regular and irregular timestamps in TimeGPT for accurate forecasting.

Irregular Timestamps

Configure confidence levels to control anomaly detection sensitivity.

Add Confidence Levels

Learn how to detect anomalies in historical time series data using confidence intervals and TimeGPT.

Historical anomaly detection

Learn how to improve anomaly detection by incorporating external factors.

How to enrich datasets with date features for historical anomaly detection.

Add date features

Intro Online (Real-Time) anomaly concepts

Online (Real-Time) Anomaly Detection

Learn how to use the detect_anomalies_online method for real-time anomaly detection in streaming time series data with TimeGPT.

Introduction to Online (Real-Time) Anomaly Detection

Learn how to refine TimeGPT’s anomaly detection process by tuning parameters for improved accuracy and alignment with specific use cases.

Adjusting the Anomaly Detection Process

Explore the differences between single and multiple variable anomaly detection approaches.

Univariate vs Multivariate Anomaly Detection

Guide to deploying and using the TimeGEN-1 model as an Azure AI endpoint.

AzureAI

Tutorial on detecting anomalies in time series data using TimeGPT.

Anomaly detection

Learn how to incorporate external variables to improve your forecasting accuracy.

Exogenous Variables Tutorial

Guide to using calendar variables and special dates to improve forecast accuracy in time series.

Holidays and special dates

Learn how to incorporate categorical variables in your TimeGPT forecasts to improve accuracy.

Adding Categorical Variables

Learn how to interpret model predictions using SHAP values to understand the impact of exogenous variables.

SHAP Values for TimeGPT and TimeGEN

Tutorials and steps for training TimeGPT for various forecasting scenarios

Training

Learn how to use the TimeGPT long-horizon model for forecasting far into the future with Nixtla client.

Learn how to generate forecasts for multiple time series simultaneously.

Adapt TimeGPT to your specific datasets for more accurate forecasts

Learn how to save, fine-tune, list, and delete TimeGPT models to optimize forecasting.

Re-using fine-tuned models

Learn how to fine-tune a model using specific loss functions, configure the Nixtla client, and evaluate performance improvements.

Fine-tuning with a Specific Loss Function

Learn how to use the finetune_depth parameter to control the extent of fine-tuning in TimeGPT models.

Controlling the Level of Fine-Tuning

Learn how to validate time series models with cross-validation and historical forecasts

Validation

Learn how to validate time series models with rolling-window cross-validation

Cross-validation

Learn how to validate TimeGPT models by comparing historical forecasts against actual data.

Historical Forecast Evaluation

Uncertainty quantification is a key component of forecasting. Learn how to quantify uncertainty in your forecasts.

Uncertainty Quantification Tutorial

Learn how to generate forecasts at different probability levels to quantify uncertainty.

Quantile Forecasts

Learn how to create robust prediction intervals to quantify uncertainty in forecasts.

Explore special topics in TimeGPT including irregular timestamps, bounded forecasts, hierarchical forecasts, missing values, and improving forecast accuracy.

Special topics

Learn how to generate forecasts with upper and lower bounds to match your business constraints.

Bounded Forecasts

Learn how to use TimeGPT for hierarchical forecasting across multiple levels.

Hierarchical Forecasting

Learn how to combine forecasts at different time frequencies to improve accuracy.

Temporal Hierarchical Forecasting with TimeGPT

Learn how to handle missing values in time series data for accurate forecasting with TimeGPT.

Missing Values

Advanced techniques to enhance TimeGPT forecast accuracy for energy and electricity.

Improve Forecast Accuracy with TimeGPT

Learn how to use TimeGPT with distributed computing frameworks for processing large datasets.

Computing at Scale Tutorial

Learn how to run TimeGPT in a distributed manner on Spark for scalable forecasting and cross-validation.

Spark

Run TimeGPT in a distributed manner using Dask for scalable forecasting.

Dask

Distribute TimeGPT forecasting jobs on Ray for scalable Python workloads.

Learn how to predict website traffic patterns using TimeGPT.

Forecasting Web Traffic

Learn how to use TimeGPT to predict Bitcoin prices and evaluate forecast accuracy.

Bitcoin Price Prediction

Learn how TimeGPT accurately predicts electricity consumption patterns

Forecasting Energy Demand

Learn how to forecast sporadic demand patterns for inventory management.

Forecasting Intermittent Demand

Learn how to use TimeGPT to forecast sales scenarios with different pricing strategies.

QUICK START

GETTING STARTED

CAPABILITIES

DEPLOYMENT

TUTORIALS

USE CASES

REFERENCE

About

Historical Anomaly Detection Quickstart

Quickstart

Steps to Detect Anomalies

Next Steps

QUICK START

GETTING STARTED

CAPABILITIES

DEPLOYMENT

TUTORIALS

USE CASES

REFERENCE

About

​Quickstart

​Steps to Detect Anomalies

​Next Steps

Quickstart

Steps to Detect Anomalies

Next Steps