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
JavaScript ai-artai-videochrome-extensionclaude-codecodexhiggsfieldjimenglocal-firstmcpmidjourneymoodboardprompt-engineeringprompt-libraryprompt-managerreference-libraryseedancevisual-bookmarksweb-clipper
"""Real Chromium library/offscreen integration, with a test-only dispatcher hook.
Native transport is covered separately; this is not native permission acceptance.
"""
import base64
import hashlib
import json
import os
import tempfile
import time
from pathlib import Path
from e2e_support import EXTENSION_DIR, extension_session
def main():
with tempfile.TemporaryDirectory(prefix='pd-agent-browser-') as tmp:
ext = Path(tmp)
for path in EXTENSION_DIR.iterdir():
if path.name != 'background.js':
(ext / path.name).symlink_to(path, target_is_directory=path.is_dir())
(ext / 'background.js').write_text((EXTENSION_DIR / 'background.js').read_text() +
'\n// Test-only transport boundary; production package has no hook.\n' +
'globalThis.agentTestDispatch = dispatchAgentOperation;\n')
with extension_session('pd-agent-library-', extension_dir=ext) as run:
page = run.open_page('library.html')
run.seed_storage(page, {'entries': []})
worker = run.context.service_workers[0]
def call(operation, data):
return worker.evaluate('([op, data]) => globalThis.agentTestDispatch(op, data)', [operation, data])
def receipt(request_id):
deadline = time.monotonic() + 30
while time.monotonic() < deadline:
result = call('get_task', {'requestId': request_id})
if result['state'] not in ('queued', 'running'):
assert result['state'] == 'completed', result
return result
time.sleep(0.05)
raise AssertionError('Agent save did not finish')
original = 'Fixture original document: 原始提示词与创作经验'.encode()
digest = hashlib.sha256(original).hexdigest()
transfer_id = 'browser-original'
call('begin_transfer', {'id': transfer_id, 'name': 'original.txt', 'mimeType': 'text/plain', 'byteSize': len(original), 'sha256': digest})
call('append_transfer', {'id': transfer_id, 'offset': 0, 'data': base64.b64encode(original).decode()})
assert call('finish_transfer', {'id': transfer_id})['state'] == 'ready'
image_bytes = base64.b64decode('iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+jW6kAAAAASUVORK5CYII=')
call('begin_transfer', {'id': 'browser-image', 'name': 'reference.png', 'mimeType': 'image/png', 'byteSize': len(image_bytes), 'sha256': hashlib.sha256(image_bytes).hexdigest()})
call('append_transfer', {'id': 'browser-image', 'offset': 0, 'data': base64.b64encode(image_bytes).decode()})
image_asset = call('finish_transfer', {'id': 'browser-image'})['assetId']
args = {'requestId': 'browser-save', 'title': 'Agent browser fixture', 'text': 'Use the original as reference.', 'kind': 'creation', 'transferIds': [transfer_id], 'filePrompts': {transfer_id: '保留附件对应提示词'}, 'creative': {'purpose':'广告参考','notes':'待审'}, 'customLabels':['人工分类'], 'sourceFacts':{'author':'fixture author'}}
args['transferIds'].append('browser-image')
args['classificationPathIds'] = ['content:reference']
args['timeNotes'] = [{'assetId': image_asset, 'startMs': 0, 'text': '保留人工画面反馈'}]
call('save_material', args)
saved = receipt(args['requestId'])
case_id = saved['result']['results'][0]['entryId']
call('save_material', args)
found = call('search', {'query': 'Agent browser fixture'})
assert found['total'] == 1, found
case = call('read_case', {'caseId': case_id})
assert case['content'] == args['text'], case
source = json.loads(call('read_case_details', {'caseId': case_id, 'part': 'source'})['content'])
assert source['provenance']['kind'] == 'creation', source
assert source['sourceFacts']['author'] == 'fixture author'
assert json.loads(call('read_case_details',{'caseId':case_id,'part':'creative'})['content']) == args['creative']
prompts = call('read_case', {'caseId': case_id, 'part': 'media_prompts'})
assert '保留附件对应提示词' in prompts['content'], prompts
asset_id = json.loads(call('read_case_details', {'caseId': case_id, 'part': 'media'})['content'])[0]['id']
media = call('read_media', {'caseId': case_id, 'assetId': asset_id})
assert base64.b64decode(media['data']) == original
assert media['sha256'] == digest
document = call('read_case', {'caseId': case_id, 'part': 'document'})
assert '原始提示词与创作经验' in document['content'], document
page.reload()
page.wait_for_function("async () => (await chrome.runtime.sendMessage({type:'GET_STATE'})).entries.length === 1")
complete = json.loads(call('read_case_details', {'caseId': case_id, 'parts': ['document', 'source', 'annotations', 'media'], 'length': 49152})['content'])
assert complete['document']['text'] == args['text']
assert complete['source']['sourceFacts']['author'] == args['sourceFacts']['author']
assert complete['annotations']['creative'] == args['creative']
assert complete['annotations']['customLabels'] == args['customLabels']
assert complete['annotations']['classification']['pathIds'] == args['classificationPathIds']
for key, value in args['timeNotes'][0].items():
assert complete['annotations']['timeNotes'][0][key] == value
assert len(complete['media']) == 2
assert base64.b64decode(call('read_media', {'caseId': case_id, 'assetId': image_asset})['data']) == image_bytes
assert page.locator('#toggle-agent-connection').count() == 1
assert page.locator('#agent-connection-help').count() == 1
run.seed_storage(page, {'capturePermissionOnboarding': {'version': 1, 'acknowledgedAt': '2026-09-12T00:00:00Z', 'clipboardIncluded': True}})
page.evaluate("() => { if (!document.querySelector('#settings-dialog').open) document.querySelector('#open-settings').click(); }")
page.locator('[data-settings-tab="general"]').click()
# Clipboard is isolated to the test; clicking still runs the actual
# instance preparation and builds the same user-facing request.
page.evaluate("() => { navigator.clipboard.writeText = async text => { window.copiedAgentRequest = text; }; }")
page.locator('#copy-agent-connection').click()
page.wait_for_function('() => !!window.copiedAgentRequest')
copied = page.evaluate('window.copiedAgentRequest')
connection = page.evaluate("async () => (await chrome.runtime.sendMessage({type:'GET_AGENT_CONNECTION'})).connection")
assert connection['instanceId'] in copied and 'connector/INSTALL.md' in copied
assert connection['enabled'] is False, connection
page.locator('#copy-agent-connection').click()
assert page.evaluate('window.copiedAgentRequest') == copied
page.locator('#copy-agent-connection').scroll_into_view_if_needed()
evidence = Path(os.environ.get('PROMPTDIRECTOR_LAB_EVIDENCE_DIR', str(ext)))
evidence.mkdir(parents=True, exist_ok=True)
page.screenshot(path=str(evidence / 'agent-onboarding.png'))
page.set_viewport_size({'width': 390, 'height': 844})
page.screenshot(path=str(evidence / 'agent-onboarding-mobile.png'))
print('PASS: actual worker, offscreen document preparation, original bytes, prompt association, search/read, retry and persistent library')
if __name__ == '__main__':
main()