Skip to content

Spot it. Save it. Free, local-first Chrome extension for images, videos and prompts. Explore 2,000+ curated cases and connect your library to AI Agents via MCP. 提示词导演

8.5 KB
Raw Blame
"""A main post, quoted video and author prompt share the real original save path."""
import json
import tempfile
from pathlib import Path
from e2e_support import EXTENSION_DIR, extension_session
from page_capture_e2e import fixture_png

WORK = 'https://x.com/director/status/123'
QUOTE = 'https://x.com/director/status/100'
CDN = 'https://video.twimg.com/quoted.mp4'
POSTER = 'https://pbs.twimg.com/amplify_video_thumb/555/img/cover.jpg'
FULL = 'Full author prompt\n' + '\n'.join(f'Shot {i}: preserve the original movement.' for i in range(90))
BYTES = (Path(__file__).parent / 'fixtures/zhipu-local-video-smoke.mp4').read_bytes()


def main():
    with tempfile.TemporaryDirectory(prefix='x-quoted-extension-') as temp:
        ext = Path(temp)
        for file in EXTENSION_DIR.iterdir():
            # Chrome's dynamically registered scripts must be real files in the
            # installed extension; a symlink is accepted by executeScript but
            # silently omitted at document_start. Packages contain real files.
            if file.name == 'x-video-observer.js':
                (ext / file.name).write_bytes(file.read_bytes())
            elif file.name != 'manifest.json':
                (ext / file.name).symlink_to(file, target_is_directory=file.is_dir())
        manifest = json.loads((EXTENSION_DIR / 'manifest.json').read_text())
        manifest['host_permissions'] += ['https://x.com/*', 'https://video.twimg.com/*', 'https://pbs.twimg.com/*']
        (ext / 'manifest.json').write_text(json.dumps(manifest))
        media = {'expanded_url': QUOTE + '/video/1', 'media_url_https': POSTER, 'video_info': {'variants': [{'content_type': 'video/mp4', 'url': CDN, 'bitrate': 10}]}}
        data = {'data': {'tweet': {'rest_id': '123', 'quoted_tweet_results': {'result': {'rest_id': '100', 'media_entities2': [media]}}}}}
        metadata = '<script type="application/x-fixture">($R=>$R[0]={kind:"GraphQLRequestStream.Completed",result:{value:' + json.dumps(json.dumps(data)) + '}})($R["tsr"])</script>'
        images = ''.join(f'<a href="{WORK}/photo/{i}"><img src="https://pbs.twimg.com/media/list-image-{i}.jpg" width="100" height="100"></a>' for i in range(1, 5))
        html = f'''<html><head><link rel="canonical" href="{WORK}"></head><body><main>
        <article data-testid="tweet"><div data-testid="User-Name"><a href="/director"><span>Director</span></a></div>
        <div data-testid="tweetText">Main post; assets and the prompt below.</div>{images}
        <div role="link" tabindex="0"><div data-testid="tweetText">Quoted original video.</div><video poster="{POSTER}" width="320" height="180"></video></div>
        <a href="/director/status/123"><time>Today</time></a></article>
        <article data-testid="tweet"><div data-testid="User-Name"><a href="/director"><span>Director</span></a></div>
        <div data-testid="tweetText">Full author prompt preview<button data-testid="tweet-text-show-more-link" onclick="this.parentElement.textContent=window.fullPrompt">Show more</button></div>
        <a href="/director/status/124"><time>Today</time></a></article>
        <article data-testid="tweet"><div data-testid="User-Name"><a href="/other">Other</a></div><div data-testid="tweetText">UNRELATED COMMENT</div><a href="/other/status/125"><time>Today</time></a></article>
        </main>{metadata}<script>window.fullPrompt={json.dumps(FULL)};fetch('/i/api/graphql/fixture').then(r=>r.json());</script></body></html>'''
        with extension_session('x-quoted-original-', extension_dir=ext) as run:
            response = {'data': [{'rest_id': '123', 'legacy': {'extended_entities': {'media': []}}, 'quoted_status_result': {'result': {'rest_id': '100', 'legacy': {'extended_entities': {'media': [media]}}}}}]}
            run.context.route('https://x.com/**', lambda route: route.fulfill(json=response) if '/i/api/' in route.request.url else route.fulfill(body=html, content_type='text/html'))
            run.context.route('https://pbs.twimg.com/**', lambda route: route.fulfill(body=fixture_png(route.request.url), content_type='image/png'))
            run.context.route(CDN, lambda route: route.fulfill(body=BYTES, content_type='video/mp4'))
            collector = run.open_page('collector.html')
            run.seed_storage(collector, {'entries': [], 'dataSafetyOnboardingSeen': True})
            worker = run.context.service_workers[0]
            worker.evaluate("""body=>{const original=fetch;globalThis.fetch=(url,options)=>String(url)==='https://x.com/director/status/123'?Promise.resolve(new Response(body,{headers:{'content-type':'text/html'}})):original(url,options)}""", html)
            source = run.context.new_page()
            source.goto(WORK)
            source.bring_to_front()
            batch = collector.evaluate("async()=>{const r=await chrome.runtime.sendMessage({type:'START_PAGE_CAPTURE',mode:'loaded'});if(!r.ok)throw Error(r.message);return r.batch;}")
            c = batch['candidates'][0]
            video = next(m for m in c['media'] if m['kind'] == 'video')
            assert video['url'] == CDN and video['originalWorkUrl'] == QUOTE and video['isQuoted'], video
            supplement = collector.evaluate("async ({item,tabId})=>chrome.runtime.sendMessage({type:'READ_PAGE_CAPTURE_SUPPLEMENT',supplement:item,sourceTabId:tabId,sourceUrl:'https://x.com/director/status/123'})", {'item':c['supplements'][0], 'tabId':batch['tabId']})
            assert supplement['ok'] and supplement['supplement']['text'] == FULL, supplement
            # Replay the unresolved selected/persisted fragment through COMMIT,
            # which previously skipped the whole-page original resolver entirely.
            video['url'] = ''
            video['variants'] = []
            c['articleDocument']['blocks'] = [b for b in c['articleDocument']['blocks'] if b.get('assetId') != video['id']]
            batch = collector.evaluate("""async ({batch,item})=>{const {appendCaptureCandidate}=await import('./capture-additions.js');const c=batch.candidates[0];const selection={candidateId:c.id,includeText:true,selectedMediaIds:c.media.map(m=>m.id),mediaDecision:'confirmed'};const block={id:item.id,kind:'paragraph',text:item.text};const r=appendCaptureCandidate(c,{id:item.id,textBlocks:[block],articleDocument:{blocks:[{id:'source',kind:'link',sourceUrl:item.sourceUrl,label:'Author prompt'},block]}},selection);return {...batch,candidates:[r.candidate],selections:[r.selection]};}""", {'batch':batch,'item':supplement['supplement']})
            # Anonymous metadata can be rejected even while the signed-in page
            # has its normal media response. Save must use the real generated
            # document-start observer in a temporary page, leaving source intact.
            worker.evaluate("""()=>{
              const original=fetch;globalThis.fetch=(url,options)=>String(url)==='https://x.com/director/status/123'?Promise.resolve(new Response('',{status:403})):original(url,options);
              const create=chrome.tabs.create.bind(chrome.tabs);
              chrome.tabs.create=async options=>{const tab=await create({...options,url:'about:blank'});await new Promise(resolve=>globalThis.fixtureTabReady=resolve);return tab;};
            }""")
            with run.context.expect_page() as opened:
                collector.evaluate("batch=>{window.savePromise=chrome.runtime.sendMessage({type:'COMMIT_PAGE_CAPTURE',batch});}", batch)
            opened.value.goto(WORK)
            worker.evaluate('()=>globalThis.fixtureTabReady()')
            result = collector.evaluate('()=>window.savePromise')
            assert result['ok'] and not result['results'][0]['warnings'], result
            collector.reload()
            saved = collector.evaluate("""async()=>{const s=await chrome.runtime.sendMessage({type:'GET_STATE'}),e=s.entries[0],a=e.mediaAssets.find(a=>a.kind==='video');return {count:s.entries.length,text:e.text,assets:e.mediaAssets,size:(await(await import('./media-store.js')).getMediaBlob(a.id)).size};}""")
            assert saved['count'] == 1 and FULL in saved['text'] and 'UNRELATED COMMENT' not in saved['text'], saved
            assert saved['size'] == len(BYTES) and len([a for a in saved['assets'] if a['kind'] == 'image']) >= 4, saved
            assert all(a['storageMode'] == 'managed' for a in saved['assets']), saved
            assert source.url == WORK and opened.value.is_closed()
            print('PASS: metadata HTTP403 recovered through normal responses in a closed temporary tab; quoted source ownership; current JSON response; author full prompt; unresolved selected fragment resolves at save; four images and original video survive reload in one case')


if __name__ == '__main__':
    main()