Compare commits

..
4 changed files with 0 additions and 307 deletions
-2
View File
@@ -2,5 +2,3 @@
node_modules/
web-ext-artifacts/
npm-debug.log*
# Listing images are generated; see docs/amo/README.md.
docs/amo/screenshots/
-116
View File
@@ -1,116 +0,0 @@
# Add-ons listing
Paste-ready text and images for the
[addons.mozilla.org listing](https://addons.mozilla.org/en-US/firefox/addon/search-with-chatgpt/).
Edit them in the AMO Developer Hub under the add-on's listing details.
Listing copy follows the copy rule in the
[options design](../superpowers/specs/2026-07-29-extension-options-design.md):
ChatGPT opens with the text *filled in*. Never say the text is sent, searched,
or asked. Keep "private", "incognito", and "secure" away from temporary chat,
and call temporary chat and web search hints.
## Branding
The name's "powered by OpenAI" and the extension's own icon follow OpenAI's
brand guidelines on purpose:
- OpenAI's logo would make unofficial software look official, so it isn't
used.
- "Powered by OpenAI" is one of the phrases the guidelines allow.
Keep both, along with the description's "unofficial" line.
## Summary
Plain text, 250 characters at most. The summary isn't Markdown, and AMO escapes
`<` and `>` in it, so `<Alt>` would show as `&lt;Alt&gt;`. The summary is the one
place that writes the shortcut without angle brackets:
```text
Ask ChatGPT from anywhere in Firefox. Type gpt in the address bar, right-click selected text, press Alt+Shift+G, or click the toolbar button. ChatGPT opens with your text filled in. Tiny, fast, no API key.
```
## Description
Since March 2025, AMO formats descriptions as
[Markdown](https://blog.mozilla.org/addons/2025/03/03/markdown/) and shows any
HTML as plain text. Plain `<Ctrl>` therefore shows as typed. Check the preview
after saving.
```markdown
Ask ChatGPT from anywhere in Firefox.
**Four ways to use it**
- **Address bar:** press <Ctrl>+L, type `gpt` and your question, like `gpt why does my cat knock things over?`, and press <Enter>.
- **Right-click:** select text on any page and choose Ask ChatGPT.
- **Keyboard:** select text and press <Alt>+<Shift>+G. With nothing selected, it opens a blank chat.
- **Toolbar:** click the Ask ChatGPT button.
ChatGPT opens with your text filled in. You may need to sign in to ChatGPT.
**Small on purpose**
The whole extension is about 56 KB, with no libraries, no server, no API key and no intermediary, and nothing is reported to us. Firefox opens chatgpt.com directly, and it never makes ChatGPT your default search engine unless you choose to.
**Make it yours**
- A simple right-click menu, or an advanced one with Explain, Summarize, Debug code and Translate.
- Your own prompt template, like `Explain simply: {query}`.
- Open ChatGPT in a new tab, a background tab or a new window.
- Ask for a temporary chat or a web search. These are hints, and ChatGPT decides whether to honour them.
- Add your own search keyword or change the shortcut in Firefox's settings.
Search with ChatGPT is an unofficial extension by Zavage Software and isn't affiliated with OpenAI.
[Source code](https://git-repos.zavage.net/zavage-software/search-with-chatgpt-powered-by-openai-extension) · [Project website](https://zavage-software.com/portfolio/search-with-ChatGPT) · [Feedback or report a bug](https://zavage-software.com/contact-us)
```
Update "about 56 KB" when the package size changes noticeably. `npm run build`
prints the ZIP's path.
## Screenshots
Upload in this order, with these captions. Each image is 1280×800, the size AMO
recommends, and carries no added text, as Mozilla's listing guide advises.
The images are not committed. Generate them as described under
[Regenerating the images](#regenerating-the-images); they land in
`docs/amo/screenshots/`, which git ignores, so the links below work once you
have.
| Image | Caption |
|---|---|
| [1-address-bar.png](screenshots/1-address-bar.png) | Type gpt and your question in the address bar. ChatGPT opens with it filled in. |
| [2-right-click.png](screenshots/2-right-click.png) | Select text on any page, right-click, and choose Ask ChatGPT. |
| [3-keyboard.png](screenshots/3-keyboard.png) | Or select text and press <Alt>+<Shift>+G. |
| [4-toolbar.png](screenshots/4-toolbar.png) | Or click the Ask ChatGPT button in the toolbar. |
| [5-advanced-menu.png](screenshots/5-advanced-menu.png) | Switch to the Advanced menu for Explain, Summarize, Debug code and Translate. |
A real screenshot of the options page in `about:addons` makes a good sixth
image: one genuine screenshot next to the drawings helps people trust them.
Suggested caption: "Every setting on one simple page."
## Regenerating the images
The images are the options page's drawings, enlarged by
[`screenshots.html`](screenshots.html), which reuses `options/options.css`. They
are generated rather than committed, like the build ZIP. When a drawing in
`options/options.html` changes, copy the change into `screenshots.html`. Then,
to produce the images, run these from the repository root:
```sh
python3 -m http.server 8766 --bind 127.0.0.1 &
profile=$(mktemp -d)
echo 'user_pref("layout.css.prefers-color-scheme.content-override", 1);' > "$profile/user.js"
for shot in 1-address-bar 2-right-click 3-keyboard 4-toolbar 5-advanced-menu; do
firefox --headless -no-remote -profile "$profile" --window-size=1280,800 \
--screenshot "$PWD/docs/amo/screenshots/$shot.png" \
"http://127.0.0.1:8766/docs/amo/screenshots.html#$shot"
done
kill %1
```
The output path must be absolute. The throwaway profile forces the light theme,
and a fresh one avoids Firefox serving an old copy of the page from its cache.
Opening `screenshots.html` without a `#fragment` shows every image in order.
-130
View File
@@ -1,130 +0,0 @@
<!doctype html>
<!-- Renders the options page drawings as 1280x800 images for the
addons.mozilla.org listing. Open with a #fragment (#1-address-bar) to
show one image; without one, every image is shown in order.
README.md in this folder explains how to regenerate them. Keep the
markup in step with the drawings in options/options.html. -->
<meta charset="utf-8" />
<title>Add-ons listing images</title>
<link rel="stylesheet" href="../../options/options.css" />
<style>
body {
max-width: none;
margin: 0;
padding: 0;
background: #ffffff;
}
.shot {
display: flex;
align-items: center;
justify-content: center;
width: 1280px;
height: 800px;
}
body:has(.shot:target) .shot:not(:target) {
display: none;
}
/* zoom re-lays the drawing out at the larger size, so text stays sharp. */
.stage {
width: 24rem;
zoom: 2.6;
}
.stage-wide {
width: 26rem;
zoom: 2.4;
}
.stage .drawing {
margin: 0;
max-width: none;
}
.stage .drawing-keycaps {
padding-top: 0.5rem;
}
</style>
<section class="shot" id="1-address-bar">
<div class="stage stage-wide">
<div class="drawing">
<div class="drawing-urlbar">
<span class="drawing-chip"><img src="../../icons/powered-by-openai-icon-128x128.png" alt="" />ChatGPT</span>
<span class="drawing-typed">why does my cat knock things over?</span>
</div>
</div>
</div>
</section>
<section class="shot" id="2-right-click">
<div class="stage">
<div class="drawing">
<p class="drawing-page">Nobody knows <mark>why cats knock things over</mark>…</p>
<div class="drawing-menu">
<div>Copy</div>
<div>Select All</div>
<hr />
<div class="drawing-menu-active"><img src="../../icons/powered-by-openai-icon-128x128.png" alt="" />Ask ChatGPT</div>
</div>
</div>
</div>
</section>
<section class="shot" id="3-keyboard">
<div class="stage">
<div class="drawing">
<p class="drawing-page">Cats sleep for <mark>sixteen hours a day</mark>.</p>
<div class="drawing-keycaps">
<kbd class="drawing-key">Alt</kbd>+<kbd class="drawing-key">Shift</kbd>+<kbd class="drawing-key">G</kbd>
</div>
</div>
</div>
</section>
<section class="shot" id="4-toolbar">
<div class="stage">
<div class="drawing">
<div class="drawing-toolbar">
<span class="drawing-address">example.com</span>
<span class="drawing-button">
<svg viewBox="0 0 16 16"><path d="M2 5.5h3a2 2 0 1 1 3.2 0h3.3v2.8a2 2 0 1 1 0 3.2v3H2z" /></svg>
</span>
<span class="drawing-button drawing-button-hover">
<img src="../../icons/powered-by-openai-icon-128x128.png" alt="" />
<svg class="drawing-pointer" viewBox="0 0 12 18"><path d="M1 1v14.5l3.6-3.4 2.6 5.6 2.2-1-2.6-5.5H11z" /></svg>
</span>
<span class="drawing-button">
<svg viewBox="0 0 16 16"><path d="M3 4.5h10M3 8h10M3 11.5h10" /></svg>
</span>
<span class="drawing-tooltip">Ask ChatGPT</span>
</div>
<p class="drawing-content"><mark>Why cats sit in every box</mark></p>
</div>
</div>
</section>
<section class="shot" id="5-advanced-menu">
<div class="stage">
<div class="drawing">
<div class="drawing-menus">
<div class="drawing-menu">
<div>Copy</div>
<div>Select All</div>
<hr />
<div class="drawing-menu-active"><img src="../../icons/powered-by-openai-icon-128x128.png" alt="" />Ask ChatGPT<span class="drawing-submenu-arrow">›</span></div>
</div>
<div class="drawing-menu drawing-submenu">
<div>Ask normally</div>
<hr />
<div class="drawing-menu-active">Explain</div>
<div>Summarize</div>
<div>Debug code</div>
<div>Translate…</div>
</div>
</div>
</div>
</div>
</section>
@@ -1,59 +0,0 @@
# Add-ons listing for 1.3
Status: implemented on `docs/24-amo-listing-add-screenshots-and-a` (#24). The
paste-ready text and captions are in [`docs/amo/`](../../amo/README.md), along
with the page that renders the images.
## Problem
The AMO listing had no screenshots, and its description covered only the `gpt`
keyword. It never mentioned right-click, the keyboard shortcut, or the toolbar
button. The listing is where newcomers decide to install and first learn how to
use the extension.
## Decisions
- **Tone: clear, with one joke.** The copy gives plain steps for newcomers. A
general cat joke ("gpt why does my cat knock things over?") replaces the old
programmer joke ("sudo make me a sammich"), and the images share the cat
theme.
- **Say it is small.** Customization without bloat is what sets this extension
apart, so the description has a "Small on purpose" section: about 56 KB, no
libraries, no server, no API key, no intermediary, and nothing reported to
us. These are the README's privacy claims; nothing new is claimed.
- **Drawings as screenshots.** The images are the options page's drawings,
enlarged and rendered by headless Firefox:
- They are part of the extension's own UI.
- They stay sharp and on brand.
- They can be regenerated after UI changes.
Real screenshots would show OS-specific menus and go stale. A real
screenshot of the options page is suggested as a sixth image, taken by the
maintainer, because one genuine screenshot builds trust in the rest.
- **AMO's guidance.** Images are 1280×800, the size AMO recommends and its
largest display size, and carry no added text. Explanations go in each image's
caption.
- **The copy rule holds.** ChatGPT opens with the text filled in; it is never
sent, searched, or asked. Temporary chat and web search are described as
hints, and "private", "incognito", and "secure" appear nowhere near temporary
chat.
- **Branding stays.** The name's "powered by OpenAI" and the extension's own
icon follow OpenAI's brand guidelines: no OpenAI logo on unofficial software,
and only the phrases the guidelines allow. The description's "unofficial"
line complements them.
- **Correction from the draft.** The description does not say ChatGPT opens "in
a new tab": `gpt` searches open in the current tab, and the other ways follow
the Open ChatGPT in setting.
## Files
- `docs/amo/README.md`: summary, description (Markdown, which AMO has used
instead of HTML since March 2025), captions in
upload order, and how to regenerate the images.
- `docs/amo/screenshots.html`: the drawing markup at 1280×800, one image per
`#fragment`, styled by `options/options.css`.
- `docs/amo/screenshots/*.png`: the five rendered images. They are generated
and git-ignored rather than committed, so the repository carries no binaries
and needs no Git LFS. The command to produce them is in the README.
`docs/` is excluded from the extension package, so none of this ships to users.