AI Chat Speed Booster

Keeps long AI chat conversations responsive by showing only recent messages first, then letting you load older ones when you need them.

Works on ChatGPT, Claude, Gemini, and any AI chat app you add to the config.

Install via official browser extension store

Browser	Version	Link
Chrome	v1.4.3	chromewebstore
Firefox	v1.4.3	addons.mozilla.org

Install

We suggest using your browser's official extension store because it updates automatically. If the extension is not available in your browser's store, either open an issue to let us know or download the extension and import it manually. Keep in mind that manually installed versions do not update automatically when a new release is published.

1. Go to Releases
2. Download the zip for your browser
3. Follow your browser guide:
   • Chrome install guide
   • Firefox install guide
   • Edge install guide
   • Safari install guide

Store listings (Chrome/Firefox) may lag because review can take time.

Build it yourself

1) Requirements

• Node.js 18+
• npm

2) Install

git clone https://github.com/Noah4ever/ai-chat-speed-booster.git
cd ai-chat-speed-booster
npm install

3) Build

Build all targets:

npm run build:all

You can also just build one target (chrome, firefox, safari, edge):

npm run build:chrome

Build output goes to dist/chrome/.

Releasing a new version

The two scripts under scripts/ handle everything between editing code and uploading zips to the stores.

1. Bump the version

npm run bump 1.4.5

Updates the version in package.json and all four browser manifests (browsers/{chrome,edge,firefox,safari}/manifest.json), stages only those five files, commits with fix: update version to 1.4.5, and creates the tag v1.4.5.

Flags:

• npm run bump 1.4.5 -- --dry — preview the changes, write nothing
• npm run bump 1.4.5 -- --no-tag — commit but skip the tag

The script refuses to run if the tag already exists, if you're already on that version, or if the version isn't semver-shaped (X.Y.Z or X.Y.Z-beta.1).

2. Package the release zips

npm run package

Builds Chrome + Firefox, then writes three zips into deploys/ (gitignored):

File	What it's for
chrome-v<version>.zip	Upload to the Chrome Web Store
firefox-v<version>.zip	Upload to Firefox AMO
firefox-source-v<version>.zip	Source archive AMO requires for review

The source zip is produced by git archive HEAD, so it always matches the last commit — bump and commit before packaging, otherwise the source archive will show the previous version's tree.

Flag: npm run package -- --skip-build re-zips whatever's already in dist/ without rebuilding.

3. Push and upload

git push && git push --tags

Then upload from deploys/:

• Chrome Web Store — upload chrome-v<version>.zip
• Firefox AMO — upload firefox-v<version>.zip, then attach firefox-source-v<version>.zip when prompted for source code. AMO also asks for release notes — paste a short summary of what changed.

4. After the stores publish

Update the version numbers in the Install via official browser extension store table at the top of this README. Store review can take days, so the listed version is allowed to lag behind the tag.

Adding a new AI chat site

All site definitions live in one file: sites.config.json.

To add a new site, add an entry to the array:

{
    "id": "mysite",
    "name": "My AI Chat",
    "hostnames": ["mysite.com"],
    "urlPatterns": ["*://mysite.com/*"],
    "selectors": {
        "messageTurn": ".message-selector",
        "scrollContainer": ".scroll-container"
    },
    "messageIdAttribute": "data-message-id"
}

Then rebuild. The build script auto-injects the URL patterns into all browser manifests. No other files need to change.

Finding the right selectors

1. Open the AI chat in your browser
2. Right-click on a message → Inspect
3. Find the repeating element wrapping each message turn → that's messageTurn
4. Find the scrollable container → that's scrollContainer
5. If there's a fallback scroll container, add scrollContainerAlt
6. If messages have a unique ID attribute, set messageIdAttribute (defaults to data-testid)

Currently supported sites

Site	Status
ChatGPT	✅ Tested
Claude	✅ Tested
Gemini	✅ Tested

PRs to add or fix site configs are welcome.

How it works

• Shows the latest messages first (default: 3)
• Hides older messages
• Adds a "Load more" button at the top to reveal older messages in batches
• Keeps the visible window capped as new messages arrive
• Caches up to 5 recent chats (LRU) for faster switching
• Trims chat/API data before rendering to reduce load time and improve responsiveness

Settings

Set these from the popup:

Setting	Default	Range
Visible messages	3	1-200
Load more batch	3	1-50
Status indicator	On	On/Off
Badge position	Top right	4 corners

Testing

Automated tests use Playwright to validate build outputs and run the real extension in headless Chromium against mock pages.

Run all tests

npm test

This builds all browser targets, validates every dist/ output, then loads the extension in Chromium and verifies it works for each configured site (60 tests, ~7s).

Individual test suites

npm run test:build       # validate dist/ outputs only
npm run test:extension   # extension tests on mock pages
npm run test:integration # live site tests (requires auth, see below)

Integration tests (live sites)

To test against real sites with your account:

1. Copy .env.example to .env and fill in your credentials
   (for ChatGPT via Google login, use your Google email/password)
2. Run npm run test:auth — a browser opens, log in to each site, press Enter
3. Run npm run test:integration

The auth profile is saved to tests/.auth-profile/ (git-ignored) and reused across runs.

Headless mode

Set HEADLESS=1 to run without a visible browser window:

HEADLESS=1 npm test

Browser support

• Chrome
• Firefox
• Edge
• Safari

Privacy

• No message content is read or sent anywhere
• No analytics or tracking
• Settings are stored locally in browser storage

Source code submission (Firefox)

This project is built from TypeScript source files and bundled with esbuild.

Build environment

• Operating systems: Linux, macOS, or Windows
• Node.js: 18 or newer
• npm: included with Node.js

Reproducible build steps (Firefox)

git clone https://github.com/Noah4ever/ai-chat-speed-booster.git
cd ai-chat-speed-booster
npm ci
npm run build:firefox

The Firefox extension output is generated in dist/firefox/. The file to load or package is dist/firefox/manifest.json.

Build script used by this project: scripts/build.mjs

Credits

Fast Mode in this project uses a fetch-interception approach that trims API responses before the app renders them. Earlier work in this area includes Speed Booster for ChatGPT by BGSN.

License

MIT