developer
developer copied to clipboard
Running example prompt in prompt.md file generates non-working chrome extension - Error: Could not load icon 'icon16.png' specified in 'icons'. Could not load manifest.
Tested several times, always with the same results.
I get no errors in the console.
Here is what I run -
modal run main.py --prompt prompt.md --model=gpt-4
Here is the result - `✓ Initialized. View app at https://modal.com/apps/ap-2X0ikxpFtqmLoE63zZSiah ✓ Created objects. ├── 🔨 Created generate_response. ├── 🔨 Created mount /mnt/nas/developer/main.py ├── 🔨 Created mount /mnt/nas/developer/utils.py ├── 🔨 Created mount /mnt/nas/developer/constants.py └── 🔨 Created generate_file. hi its me, 🐣the smol developer🐣! you said you wanted: a Chrome Manifest V3 extension that reads the current page, and offers a popup UI that has the page title+content and a textarea for a prompt (with a default value we specify). When the user hits submit, it sends the page title+content to the Anthropic Claude API along with the up to date prompt to summarize it. The user can modify that prompt and re-send the prompt+content to get another summary view of the content.
- Only when clicked:
- it injects a content script
content_script.js
on the currently open tab, and accesses the titlepageTitle
and main content (innerText)pageContent
of the currently open page (extracted via an injected content script, and sent over using astorePageContent
action) - in the background, receives the
storePageContent
data and stores it - only once the new page content is stored, then it pops up a full height window with a minimalistic styled html popup
- in the popup script
- the popup should display a 10px tall rounded css animated red and white candy stripe loading indicator
loadingIndicator
, while waiting for the anthropic api to return- with the currently fetching page title and a running timer in the center showing time elapsed since call started
- do not show it until the api call begins, and hide it when it ends.
- retrieves the page content data using a
getPageContent
action (and the background listens for thegetPageContent
action and retrieves that data) and displays the title at the top of the popup - check extension storage for an
apiKey
, and if it isn't stored, asks for an API key to Anthropic Claude and stores it. - at the bottom of the popup, show a vertically resizable form that has:
- a 2 line textarea with an id and label of
userPrompt
-
userPrompt
has a default value ofdefaultPrompt = `Please provide a detailed, easy to read HTML summary of the given content`; ```js
-
- a 4 line textarea with an id and label of
stylePrompt
-
stylePrompt
has a default value ofdefaultStyle = `Respond with 3-4 highlights per section with important keywords, people, numbers, and facts bolded in this HTML format: <h1>{title here}</h1> <h3>{section title here}</h3> <details> <summary>{summary of the section with <strong>important keywords, people, numbers, and facts bolded</strong> and key quotes repeated}</summary> <ul> <li><strong>{first point}</strong>: {short explanation with <strong>important keywords, people, numbers, and facts bolded</strong>}</li> <li><strong>{second point}</strong>: {same as above}</li> <li><strong>{third point}</strong>: {same as above}</li> <!-- a fourth point if warranted --> </ul> </details> <h3>{second section here}</h3> <p>{summary of the section with <strong>important keywords, people, numbers, and facts bolded</strong> and key quotes repeated}</p> <details> <summary>{summary of the section with <strong>important keywords, people, numbers, and facts bolded</strong> and key quotes repeated}</summary> <ul> <!-- as many points as warranted in the same format as above --> </ul> </details> <h3>{third section here}</h3> <!-- and so on, as many sections and details/summary subpoints as warranted --> With all the words in brackets replaced by the summary of the content. sanitize non visual HTML tags with HTML entities, so <template> becomes
-
- a 2 line textarea with an id and label of
- the popup should display a 10px tall rounded css animated red and white candy stripe loading indicator
- it injects a content script
<template> but stays the same. Only draw from the source content, do not hallucinate. Finally, end with other questions that the user might want answered based on this source content:
<hr>
<h2>Next prompts</h2>
<ul>
<li>{question 1}</li>
<li>{question 2}</li>
<li>{question 3}</li>
</ul>`;
```js
- and in the last row, on either side,
- and a nicely styled submit button with an id of `sendButton` (tactile styling that "depresses" on click)
- only when `sendButton` is clicked, calls the Anthropic model endpoint https://api.anthropic.com/v1/complete with:
- append the page title
- append the page content
- add the prompt which is a concatenation of
```js
finalPrompt = `Human: ${userPrompt} \n\n ${stylePrompt} \n\n Assistant:`
```
- and use the `claude-instant-v1` model (if `pageContent` is <70k words) or the `claude-instant-v1-100k` model (if more)
- requesting max tokens = the higher of (25% of the length of the page content, or 750 words)
- if another submit event is hit while the previous api call is still inflight, cancel that and start the new one
- renders the Anthropic-generated result at the top of the popup in a div with an id of `content`
Important Details:
-
It has to run in a browser environment, so no Nodejs APIs allowed.
-
the return signature of the anthropic api is curl https://api.anthropic.com/v1/complete
-H "x-api-key: $API_KEY"
-H 'content-type: application/json'
-d '{ "prompt": "\n\nHuman: Tell me a haiku about trees\n\nAssistant: ", "model": "claude-v1", "max_tokens_to_sample": 1000, "stop_sequences": ["\n\nHuman:"] }' {"completion":" Here is a haiku about trees:\n\nSilent sentinels, \nStanding solemn in the woods,\nBranches reaching sky.","stop":"\n\nHuman:","stop_reason":"stop_sequence","truncated":false,"log_id":"f5d95cf326a4ac39ee36a35f434a59d5","model":"claude-v1","exception":null} -
in the string prompt sent to Anthropic, first include the page title and page content, and finally append the prompt, clearly vertically separated by spacing.
-
if the Anthropic api call is a 401, handle that by clearing the stored anthropic api key and asking for it again.
-
add styles to make sure the popup's styling follows the basic rules of web design, for example having margins around the body, and a system font stack.
-
style the popup body with but insist on body margins of 16 and a minimum width of 400 and height of 600.
debugging notes
inside of background.js, just take the getPageContent response directly
chrome.runtime.onMessage.addListener((request, sender, sendResponse) => {
if (request.action === 'storePageContent') {
// dont access request.pageContent
chrome.storage.local.set({ pageContent: request }, () => {
sendResponse({ success: true });
});
} else if (request.action === 'getPageContent') {
chrome.storage.local.get(['pageContent'], (result) => {
// dont access request.pageContent
sendResponse(result);
});
}
return true;
});
inside of popup.js, Update the function calls to requestAnthropicSummary
in popup.js
to pass the apiKey
:
chrome.storage.local.get(['apiKey'], (result) => {
const apiKey = result.apiKey;
requestAnthropicSummary(defaultPrompt, apiKey);
});
sendButton.addEventListener('click', () => {
chrome.storage.local.get(['apiKey'], (result) => {
const apiKey = result.apiKey;
requestAnthropicSummary(userPrompt.value, apiKey);
});
});
in popup.js
, store the defaultPrompt at the top level.
also, give a HTML format to the anthropic prompt
89 tokens in prompt: You are an AI developer who is trying to write a p... 1785 tokens in prompt: a Chrome Manifest V3 extension that reads the curr... [ "manifest.json", "background.js", "content_script.js", "popup.html", "popup.js", "styles.css" ] 145 tokens in prompt: You are an AI developer who is trying to write a p... 1785 tokens in prompt: a Chrome Manifest V3 extension that reads the curr... Shared dependencies:
-
Exported variables:
-
defaultPrompt
-
defaultStyle
-
finalPrompt
-
-
Data schemas:
-
storePageContent
action data -
getPageContent
action data
-
-
ID names of DOM elements:
-
loadingIndicator
-
userPrompt
-
stylePrompt
-
sendButton
-
content
-
-
Message names:
-
storePageContent
-
getPageContent
-
-
Function names:
-
requestAnthropicSummary
shared_dependencies.md Shared dependencies:
-
-
Exported variables:
-
defaultPrompt
-
defaultStyle
-
finalPrompt
-
-
Data schemas:
-
storePageContent
action data -
getPageContent
action data
-
-
ID names of DOM elements:
-
loadingIndicator
-
userPrompt
-
stylePrompt
-
sendButton
-
content
-
-
Message names:
-
storePageContent
-
getPageContent
-
-
Function names:
-
requestAnthropicSummary
2032 tokens in prompt: You are an AI developer who is trying to write a p... 1952 tokens in prompt: We have broken up the program into per-file g... 2032 tokens in prompt: You are an AI developer who is trying to write a p... 1952 tokens in prompt: We have broken up the program into per-file g... 2032 tokens in prompt: You are an AI developer who is trying to write a p... 1954 tokens in prompt: We have broken up the program into per-file g... 2032 tokens in prompt: You are an AI developer who is trying to write a p... 1952 tokens in prompt: We have broken up the program into per-file g... 2032 tokens in prompt: You are an AI developer who is trying to write a p... 1952 tokens in prompt: We have broken up the program into per-file g... content_script.js // content_script.js
-
function getPageTitleAndContent() { const pageTitle = document.title; const pageContent = document.body.innerText; return { pageTitle, pageContent }; }
chrome.runtime.sendMessage( { action: 'storePageContent', ...getPageTitleAndContent() }, (response) => { if (response.success) { console.log('Page content stored successfully'); } else { console.error('Failed to store page content'); } } ); 2032 tokens in prompt: You are an AI developer who is trying to write a p... 1952 tokens in prompt: We have broken up the program into per-file g... background.js chrome.runtime.onInstalled.addListener(() => { chrome.action.onClicked.addListener((tab) => { chrome.scripting.executeScript({ target: { tabId: tab.id }, files: ['content_script.js'], }); }); });
chrome.runtime.onMessage.addListener((request, sender, sendResponse) => { if (request.action === 'storePageContent') { chrome.storage.local.set({ pageContent: request }, () => { sendResponse({ success: true }); chrome.action.setPopup({ popup: 'popup.html' }, () => { chrome.action.openPopup(); }); }); } else if (request.action === 'getPageContent') { chrome.storage.local.get(['pageContent'], (result) => { sendResponse(result); }); } return true; }); manifest.json { "manifest_version": 3, "name": "Anthropic Claude Summarizer", "version": "1.0.0", "description": "A Chrome extension that summarizes web pages using the Anthropic Claude API.", "permissions": ["activeTab", "storage"], "action": { "default_popup": "popup.html", "default_icon": { "16": "icon16.png", "48": "icon48.png", "128": "icon128.png" } }, "background": { "service_worker": "background.js" }, "content_scripts": [ { "matches": ["<all_urls>"], "js": ["content_script.js"], "run_at": "document_idle" } ], "icons": { "16": "icon16.png", "48": "icon48.png", "128": "icon128.png" } } popup.html
#loadingIndicator { position: absolute; top: 50%; left: 50%; transform: translate(-50%, -50%); width: 100%; height: 10px; overflow: hidden; background-color: #f0f0f0; border-radius: 5px; }
#loadingIndicator:before { content: ""; position: absolute; top: 0; left: 0; width: 50%; height: 100%; background: linear-gradient(45deg, #f06, transparent); animation: loading 1s infinite; }
@keyframes loading { 0% { transform: translateX(-100%); } 100% { transform: translateX(100%); } }
textarea { resize: vertical; }
#sendButton { background-color: #4CAF50; border: none; color: white; padding: 8px 16px; text-align: center; text-decoration: none; display: inline-block; font-size: 16px; margin: 4px 2px; cursor: pointer; border-radius: 4px; transition: background-color 0.3s, transform 0.3s; }
#sendButton:hover { background-color: #45a049; }
#sendButton:active { transform: translateY(2px); } popup.js document.addEventListener('DOMContentLoaded', () => { const loadingIndicator = document.getElementById('loadingIndicator'); const userPrompt = document.getElementById('userPrompt'); const stylePrompt = document.getElementById('stylePrompt'); const sendButton = document.getElementById('sendButton'); const content = document.getElementById('content');
const defaultPrompt = 'Please provide a detailed, easy to read HTML summary of the given content'; const defaultStyle = `Respond with 3-4 highlights per section with important keywords, people, numbers, and facts bolded in this HTML format:
{title here}
{section title here}
{summary of the section with important keywords, people, numbers, and facts bolded and key quotes repeated}
- {first point}: {short explanation with important keywords, people, numbers, and facts bolded}
- {second point}: {same as above}
- {third point}: {same as above}
{second section here}
{summary of the section with important keywords, people, numbers, and facts bolded and key quotes repeated}
{summary of the section with important keywords, people, numbers, and facts bolded and key quotes repeated}
{third section here}
With all the words in brackets replaced by the summary of the content. sanitize non visual HTML tags with HTML entities, so becomes <template> but stays the same. Only draw from the source content, do not hallucinate. Finally, end with other questions that the user might want answered based on this source content:
Next prompts
- {question 1}
- {question 2}
- {question 3}
userPrompt.value = defaultPrompt; stylePrompt.value = defaultStyle;
function requestAnthropicSummary(prompt, apiKey) { loadingIndicator.style.display = 'block';
chrome.runtime.sendMessage({ action: 'getPageContent' }, (response) => {
const { pageTitle, pageContent } = response.pageContent;
const finalPrompt = `Human: ${prompt} \n\n ${stylePrompt.value} \n\n Assistant:`;
const model = pageContent.length < 70000 ? 'claude-instant-v1' : 'claude-instant-v1-100k';
const maxTokens = Math.max(Math.ceil(pageContent.length * 0.25), 750);
fetch('https://api.anthropic.com/v1/complete', {
method: 'POST',
headers: {
'x-api-key': apiKey,
'content-type': 'application/json',
},
body: JSON.stringify({
prompt: `${pageTitle}\n\n${pageContent}\n\n${finalPrompt}`,
model,
max_tokens_to_sample: maxTokens,
stop_sequences: ['\n\nHuman:'],
}),
})
.then((res) => {
if (res.status === 401) {
chrome.storage.local.remove('apiKey');
throw new Error('Invalid API key');
}
return res.json();
})
.then((data) => {
content.innerHTML = data.completion;
loadingIndicator.style.display = 'none';
})
.catch((error) => {
console.error(error);
loadingIndicator.style.display = 'none';
});
});
}
chrome.storage.local.get(['apiKey'], (result) => { if (!result.apiKey) { const apiKey = prompt('Please enter your Anthropic Claude API key:'); chrome.storage.local.set({ apiKey }, () => { requestAnthropicSummary(defaultPrompt, apiKey); }); } else { requestAnthropicSummary(defaultPrompt, result.apiKey); } });
sendButton.addEventListener('click', () => { chrome.storage.local.get(['apiKey'], (result) => { requestAnthropicSummary(userPrompt.value, result.apiKey); }); }); }); ✓ App completed.`
Beautiful right? No errors.
But when I go into the generated
directory, I see these files -
background.js content_script.js manifest.json popup.html popup.js shared_dependencies.md styles.css
And in manifest.json
, I see this -
{ "manifest_version": 3, "name": "Anthropic Claude Summarizer", "version": "1.0.0", "description": "A Chrome extension that summarizes web pages using the Anthropic Claude API.", "permissions": ["activeTab", "storage"], "action": { "default_popup": "popup.html", "default_icon": { "16": "icon16.png", "48": "icon48.png", "128": "icon128.png" } }, "background": { "service_worker": "background.js" }, "content_scripts": [ { "matches": ["<all_urls>"], "js": ["content_script.js"], "run_at": "document_idle" } ], "icons": { "16": "icon16.png", "48": "icon48.png", "128": "icon128.png" } }
Those icon files, icon16.png, icon48.png, and icon128.png don't exist. So when I try to load the plugin extension in chrome, I get the error - Could not load icon 'icon16.png' specified in 'icons'. Could not load manifest.
I'm starting to think however that maybe I'm missing the point of smol dev. The point is to get you 90+% of the way there, not actually build a 100% working app right?
I mean, I can easily add my own icons in this case.
So what I did was just copy the 3 image icons from the "chrome example" folder to the "generated" folder, and the extension worked. So in reality, the example prompt made working code, and I just had to supply the images myself.
If that is what the goal of smol developer, I'm perfectly happy with that.
I would just highly recommend you make this more clear in the readme documentation that when you run the example you tell us to run to test if smol dev is working or not, you need to take the extra step of supplying the chrome extension image files.
@swyxio, please confirm my assumptions here, and I'll close this.
@Hisma yes, these assumptions are correct! it was always designed to co-design a codebase, and eventually be cleanly weaned off.