ChatGPT Long Conversation Helper

A privacy-first Tampermonkey userscript that helps navigate long ChatGPT conversations by collapsing and expanding questions and answers locally in the browser.

This is a third-party local userscript. It is not an official OpenAI or ChatGPT feature.

Why this project

Long ChatGPT conversations are useful for complex technical workflows, but they become difficult to review after many rounds.

This is especially common in workflows such as:

• technical writing planning;
• code review and debugging discussions;
• API and tooling design notes;
• documentation drafting;
• AI-assisted engineering workflows;
• long-form project planning.

The first version of this project focuses on one practical problem: reducing visual noise in long conversations without exporting, uploading, or externally processing conversation content.

The first implementation started from long assistant replies, then added question collapsing because long prompts can also make technical conversations difficult to scan.

MVP scope

Version v0.1.1 focuses on local collapse and expand behavior.

It supports:

• collapsing user questions;
• collapsing assistant answers;
• showing a three-line preview;
• adding a fade mask near the third preview line;
• collapsing all visible conversation messages;
• expanding all visible conversation messages;
• hiding the full control panel into a compact LCH launcher;
• remembering local collapsed / expanded state with localStorage.

Future versions may add lightweight labels for marking important or secondary messages, but the MVP intentionally avoids extra UI complexity.

Features

• Adds a local collapse / expand button to ChatGPT user questions and assistant answers.
• Keeps a three-line preview when a message is collapsed.
• Adds a subtle fade effect near the third preview line.
• Provides bottom-right global controls:
  • Collapse all
  • Expand all
  • Hide controls
• Converts the hidden global panel into a compact LCH launcher button.
• Uses MutationObserver to detect dynamically added messages.
• Uses CSS injection for local visual changes.
• Uses localStorage to remember local collapsed / expanded state and local panel visibility.
• Runs only on https://chatgpt.com/*.
• Requires no backend server, build tool, npm package, or external dependency.

What it does not do

This project is intentionally limited.

• It does not send, upload, collect, or transmit conversation content.
• It does not call the ChatGPT API.
• It does not automate sending messages.
• It does not scrape or export conversations.
• It does not bypass ChatGPT limits or protections.
• It does not read cookies, tokens, account data, payment data, or settings.
• It only modifies the local browser view.
• It stores only local UI state in localStorage.

Installation

1. Install Tampermonkey

Install Tampermonkey from your browser's official extension store.

After installation, pin the Tampermonkey icon to the browser toolbar if it is hidden behind the browser extension menu.

2. Enable userscript execution in Chrome if required

Some Chrome / Chromium-based browser versions require userscript execution to be explicitly allowed before Tampermonkey scripts can run.

Open:

chrome://extensions/

Then:

1. Find Tampermonkey.
2. Open Details.
3. Enable Allow User Scripts if the option is shown.
4. If Chrome asks for Developer Mode before allowing userscripts, enable Developer Mode and then enable userscript execution.
5. Refresh https://chatgpt.com/ after changing the setting.

If this step is missed, the script may be enabled in Tampermonkey but still not appear on the ChatGPT page.

3. Create a new userscript

Open the Tampermonkey dashboard and create a new script.

Usually:

Tampermonkey icon → Dashboard → Create a new script

4. Paste the script

Copy the complete content of:

chatgpt-long-conversation-helper.user.js

Delete the default Tampermonkey template and paste the script into the editor.

5. Save and enable the script

Save the userscript with:

Ctrl + S

Then confirm the script is enabled in Tampermonkey.

6. Open ChatGPT

Open a long ChatGPT conversation at:

https://chatgpt.com/

The script should add local collapse / expand controls to visible questions and answers.

Usage

Collapse one question or answer

Click the button above a message:

Collapse question

or:

Collapse answer

The message will shrink to a three-line preview.

Expand one question or answer

Click:

Expand question

or:

Expand answer

The full message will become visible again.

Collapse all messages

Use the floating control panel in the bottom-right corner and click:

Collapse all

Expand all messages

Click:

Expand all

Hide and reopen global controls

Click:

Hide controls

The full panel will collapse into a compact LCH launcher near the lower-right area.

Click:

LCH

to reopen the full control panel.

Screenshots

Before: long conversation without collapsed messages

Single message control

After: multiple messages collapsed

Global controls

Compact LCH launcher

Privacy

This script is designed as a local browser UI enhancement.

It does not send conversation content anywhere.

The only persisted data is local UI state stored in localStorage, scoped by the current URL path, message role, and message index.

Example state:

clch:v0.1.1:/c/example-conversation:assistant:4:collapsed = 1

The script does not store message text.

For more details, see:

docs/privacy.md

Technical design

The script uses a simple browser-side architecture:

1. Tampermonkey metadata limits the script to https://chatgpt.com/*.
2. A selector strategy finds visible user and assistant message containers.
3. Each message receives a local collapse / expand toolbar.
4. The script does not re-parent or re-wrap ChatGPT message content.
5. CSS controls preview height, overflow, and fade effect.
6. A floating global control panel can collapse or expand all processed messages.
7. The full panel can be hidden into a compact LCH launcher.
8. A MutationObserver detects newly inserted messages.
9. localStorage stores only collapsed / expanded state and local panel visibility.

Selector strategy

The script prefers shallow role-based selectors such as:

[data-message-author-role="user"]
[data-message-author-role="assistant"]

It avoids deep class-name chains because ChatGPT's frontend DOM can change.

The selector configuration is grouped near the top of the userscript so it can be adjusted later if the ChatGPT page structure changes.

Code-level privacy review

The userscript is intentionally small and local. In the current codebase, there is no use of:

fetch
XMLHttpRequest
navigator.sendBeacon
WebSocket
eval
document.cookie
@require
remote script URLs
analytics SDKs
telemetry endpoints

The main browser APIs used are:

querySelectorAll
MutationObserver
localStorage
classList
addEventListener

This keeps the project aligned with its privacy-first scope: local UI changes only.

Limitations

This is a best-effort UI enhancement.

Known limitations:

• ChatGPT DOM changes may break selectors.
• Streaming replies may not always be processed immediately.
• Collapse state is index-based and may shift if the conversation structure changes.
• Refresh-state recovery is useful but not guaranteed to be perfect.
• The script is tested manually, not against an official ChatGPT extension API.

For more details, see:

docs/limitations.md

Troubleshooting

If buttons do not appear:

1. Confirm the script is enabled in Tampermonkey.
2. Confirm the page URL starts with https://chatgpt.com/.
3. In Chrome, confirm Allow User Scripts is enabled for Tampermonkey if the option is shown.
4. Save the script and refresh the ChatGPT page.
5. Open a long conversation with several user questions and assistant answers.
6. Check the browser console for [CLCH] warnings.

For more details, see:

docs/troubleshooting.md

Roadmap

v0.1.1

• Collapse and expand questions.
• Collapse and expand answers.
• Three-line preview with fade effect.
• Global collapse / expand controls.
• Compact LCH launcher after hiding controls.
• MutationObserver support.
• Local collapsed / expanded state with localStorage.
• Privacy-first documentation.
• Reduced layout risk by avoiding message DOM re-parenting.

Possible v0.2.0

• Lightweight message labels.
• Color labels for important / secondary messages.
• Configurable preview line count.
• Improved selector fallback.
• Optional keyboard shortcuts.
• Better dark-mode visual tuning.

Not planned for the MVP

• Chrome Extension.
• Edge Extension.
• Cloud sync.
• API integration.
• Conversation export.
• AI summarization.
• User accounts.
• Backend server.

Manual testing and FAQ

For detailed local testing steps, see:

docs/manual-test-plan.md

For installation and usage questions, see:

docs/faq.md

For public screenshot prompts, see:

docs/screenshot-test-prompts.md

Related writing

• Build a Privacy-First Tampermonkey Script for Long ChatGPT Conversations
• AI-Assisted Development Is Not Autopilot

License

MIT License.

See:

LICENSE