ChatGPT Conversation Exporter

A CLI-only Laravel 12 application that turns the official ChatGPT data export (specifically conversations.json and the accompanying file_*.jpeg assets) into reader-friendly artifacts. It streams huge JSON dumps (>200 MB) without loading them fully into memory, asks whether to export conversations automatically or one-by-one, copies image assets (plus thumbnails), and can write each conversation as Markdown, PDF, or both.

Features

• Streaming parser powered by halaxa/json-machine keeps memory usage constant, even with massive exports.
• Approvals: choose automatic export or confirm each conversation interactively.
• Safe filenames with collision handling and optional attachment copying/reference modes.
• Image handling: copies relevant file_*.jpeg assets beside each transcript, generates thumbnails, and rewrites Markdown/PDF references.
• Dual outputs: Markdown, PDF (rendered via headless Chrome through Browsershot + Puppeteer/Playwright), or both; Markdown is converted through CommonMark so PDFs match the same transcript content.

Requirements

• PHP 8.2+
• Composer
• GD or Imagick extension (Intervention Image uses GD by default)
• Node.js 18+ with npm/yarn plus either Puppeteer or Playwright so Browsershot can talk to a headless Chrome/Chromium binary.

Installation

git clone https://github.com/your-github-username/chatgpt-md-exporter.git
cd chatgpt-md-exporter
composer install

Headless PDF Engine Setup

Browsershot proxies PDF generation to Node, so make sure a headless browser is available:

# Puppeteer (downloads Chromium automatically)
npm install -g puppeteer

# or Playwright
npm install -g playwright
npx playwright install chromium

If Node/npm or Chrome/Chromium live outside your PATH, set the optional overrides in .env (see .env.example) such as BROWSERSHOT_NODE_BINARY, BROWSERSHOT_NPM_BINARY, BROWSERSHOT_CHROME_PATH, or BROWSERSHOT_DISABLE_SANDBOX=true.

Command Overview

All exporting is handled by a single Artisan command:

php artisan conversations:export

Required Argument

Argument	Description
file	Absolute or relative path to conversations.json inside the ChatGPT export bundle.

Common Options

Option	Description
--output=	Target directory for Markdown/PDF files. Defaults to storage/app/conversations. The exporter will create the folder if it does not exist.
--assets=	Directory containing the exported attachment files (file_*.jpeg, etc.). Defaults to the folder that contains conversations.json.
--mode=	auto exports everything with no prompts; manual asks before each conversation. If omitted, you’ll be prompted interactively.
--asset-mode=	copy (default) copies images into the transcript folder; reference keeps links pointing to the original export directory.
--thumbnail-width=	Width (in px) of generated thumbnails used in Markdown/PDF (default 512). Use 0 to skip thumbnails entirely.
--format=	markdown, pdf, or both (comma-separated also supported, e.g. markdown,pdf). Default is markdown.

Usage Examples

Export everything into Markdown plus thumbnails, approving each conversation manually:

php artisan conversations:export \
  ../chatgpt-export/conversations.json \
  --assets=../chatgpt-export \
  --output=storage/app/markdown \
  --mode=manual \
  --thumbnail-width=320

Export both Markdown and PDF while referencing the original assets (no copies):

php artisan conversations:export \
  /path/to/conversations.json \
  --asset-mode=reference \
  --format=both

Generate PDF-only output with thumbnails disabled (useful when you already have attachments somewhere else):

php artisan conversations:export \
  /path/to/conversations.json \
  --format=pdf \
  --thumbnail-width=0

Output Layout

For each conversation:

• Markdown file: <slug>.md
• Optional PDF: <slug>.pdf
• Attachments: <slug>_assets/ containing copied images.
• Thumbnails (if enabled): <slug>_assets/thumbnails/.

Filenames are deduplicated so re-running the command won’t overwrite existing exports unless they share the exact slug and extension.

Tips

• Keep the original conversations.json and attachment directory in the same relative layout; the exporter guesses image paths from the JSON pointer IDs.
• When processing very large exports in manual mode, consider --mode=auto first to inspect the summary, then re-run in manual mode if needed.
• PDFs use a simple built-in stylesheet. If you want fully branded output later, you can extend convertMarkdownToHtml() with your own HTML/CSS template.

Troubleshooting

• Missing images: ensure --assets points to the folder that contains the file_*.jpeg files. If they’re missing, Markdown will display the pointer ID instead.
• Thumbnail errors: the command logs warnings if GD/Imagick cannot read an image. Thumbnails are skipped for those files, but the full image link remains.
• PDF rendering issues: Browsershot needs Chromium plus Node. Ensure Puppeteer/Playwright downloaded a browser and adjust the BROWSERSHOT_* env vars if Chrome lives outside your PATH. When running in restricted environments (CI, containers), you may also need to set BROWSERSHOT_DISABLE_SANDBOX=true.

Happy exporting! 🚀