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
"""Reference choices through real UI and durable jobs; a local provider records exact payloads."""
import json
import os
import tempfile
import threading
from pathlib import Path
from playwright.sync_api import expect
from e2e_support import ai_configuration_fixture, extension_session, wait_for_async_condition
from composer_video_dialogue_e2e import streaming_provider
from zhipu_glm_analysis_e2e import extension_with_local_provider_permission
from compatible_composer_stream_e2e import PNG
from composer_e2e_support import set_composer_reference_media, set_composer_direction
def main():
requests = []
release = threading.Event()
release.set()
with streaming_provider(requests, release) as origin, extension_with_local_provider_permission(origin) as directory, extension_session('reference-inputs-', extension_dir=directory) as run:
setup = run.open_page('collector.html')
run.seed_storage(setup, {'entries': [], **ai_configuration_fixture(providers={'custom-text': {
'apiKey':'fixture-key', 'endpoint':origin+'/chat/completions','protocol':'chat_completions','consent':True,
'models':{'creativePlanning':'unknown-multimodal'},
'discoveredModels':[{'id':'unknown-multimodal','tasks':['creativePlanning'],'inputModalities':['text'],'outputModalities':['text'],'status':'available'}]
}, 'custom-media': {'apiKey':'saved-vision-key','endpoint':origin+'/chat/completions','protocol':'chat_completions','consent':True,
'models':{'imageAnalysis':'saved-understanding-model'}}}, assignments={'creativePlanning':{'providerId':'custom-text','model':'unknown-multimodal'}})})
setup.evaluate('''async png=>{
const {saveMediaBlob}=await import('./media-store.js');
const {createComposerSession}=await import('./composer.js');
await saveMediaBlob('image',new Blob([Uint8Array.from(atob(png),c=>c.charCodeAt(0))],{type:'image/png'}));
await saveMediaBlob('video',new Blob(['synthetic video'],{type:'video/mp4'}));
const session=createComposerSession({id:'reference-inputs',targetType:'video',routeMode:'analyze_materials',
aiProfile:{serviceId:'custom-text',model:'unknown-multimodal'},imageReferenceMode:'text_only',videoReferenceMode:'text_only',
referenceSnapshots:[{entryId:'image-case',alias:'@参考1',referenceKind:'prompt',referenceText:'图片原始提示词',imageRefs:[{visualId:'image'}]},
{entryId:'video-case',alias:'@参考2',referenceKind:'video_sources',referenceText:'视频原始提示词和逆推文字',assetRefs:[{assetId:'video',kind:'video',mimeType:'video/mp4'}]}]});
await chrome.runtime.sendMessage({type:'UPSERT_COMPOSER_SESSION',session});
}''', PNG)
page=run.open_page('composer.html?session=reference-inputs', wait_until='networkidle')
expect(page.locator('#composer-send-images')).not_to_be_checked()
expect(page.locator('#composer-send-videos')).not_to_be_checked()
direction = page.locator('#composer-direction-trigger')
expect(direction).to_have_attribute('title', '创作方向:视频。影响提示词和可选参考,不会直接生成成品。')
direction.focus(); direction.press('Enter')
expect(page.locator('#composer-direction input[value="image"]')).to_be_visible()
page.locator('#composer-direction input[value="image"]').focus()
page.keyboard.press('Space')
expect(page.locator('#composer-direction')).not_to_have_attribute('open', '')
expect(direction).to_be_focused()
expect(direction).to_have_attribute('title', '创作方向:图片。影响提示词和可选参考,不会直接生成成品。')
set_composer_direction(page, 'video')
page.reload()
expect(page.locator('#composer-direction-icon')).to_have_attribute('href', 'assets/ui-icons.svg#icon-video')
assert len(requests) == 0
page.locator('#composer-model-trigger').click()
expect(page.locator('#composer-model-dynamic')).to_contain_text('saved-understanding-model')
page.locator('#composer-model-trigger').click()
for images, videos in [(False,False),(True,False),(False,True),(True,True)]:
set_composer_reference_media(page, images=images, videos=videos)
page.reload()
expect(page.locator('#composer-send-images')).to_be_checked(checked=images)
expect(page.locator('#composer-send-videos')).to_be_checked(checked=videos)
count=len(requests)
page.locator('#composer-instruction').fill('总结已选参考内容')
page.locator('#composer-action').click()
wait_for_async_condition(page,"async()=>{const s=await chrome.storage.local.get('creativeJobs');return s.creativeJobs?.items?.at(-1)?.status==='completed'}")
expect(page.locator('.composer-message.analysis')).to_have_count(count+1)
assert len(requests)==count+1
body=requests[-1]
content=body['messages'][-1]['content']
assert sum(p['type']=='image_url' for p in content)==int(images), body
assert sum(p['type']=='video_url' for p in content)==int(videos), body
assert '视频原始提示词和逆推文字' in json.dumps(body,ensure_ascii=False)
assert '图片原始提示词' in json.dumps(body,ensure_ascii=False)
screenshots = Path(os.environ.get('PROMPTDIRECTOR_E2E_ARTIFACT_DIR', Path(tempfile.gettempdir()) / 'promptdirector-reference-inputs'))
screenshots.mkdir(parents=True, exist_ok=True)
page.emulate_media(reduced_motion='reduce')
for width in [1280,900,640,390]:
page.set_viewport_size({'width':width,'height':900})
if width <= 640:
if 'nav-open' in (page.locator('#composer-shell').get_attribute('class') or ''): page.locator('#composer-nav-close').click()
page.wait_for_function("()=>document.querySelector('#composer-nav').getBoundingClientRect().right<=0")
assert page.evaluate('()=>document.documentElement.scrollWidth<=innerWidth+1')
expect(page.locator('.composer-input-tools #composer-reference-inputs')).to_be_visible()
expect(page.locator('#composer-image-input-status, #composer-image-input-model')).to_have_count(0)
assert page.locator('#composer-reference-inputs').bounding_box()['y'] >= page.locator('#composer-instruction').bounding_box()['y'] + page.locator('#composer-instruction').bounding_box()['height']
for theme in ['dark','light']:
page.evaluate('(theme)=>document.documentElement.dataset.theme=theme',theme)
page.screenshot(path=str(screenshots/f'{theme}-{width}.png'),full_page=True,animations='disabled')
page.locator('.composer-input-box').screenshot(path=str(screenshots/f'toolbar-{theme}-{width}.png'),animations='disabled')
for menu_id in ['composer-direction','composer-reference-inputs','composer-options']:
menu=page.locator('#'+menu_id)
trigger=menu.locator('summary')
trigger.click()
page.wait_for_function("""id=>{
const m=document.getElementById(id), p=m.querySelector('.composer-options-panel');
const r=p.getBoundingClientRect(), a=m.querySelector('summary').getBoundingClientRect();
return m.open && r.left>=0 && r.right<=document.querySelector('#composer-shell').getBoundingClientRect().right+1 && (id==='composer-options' || r.bottom<=a.top);
}""",arg=menu_id)
page.evaluate('async()=>{await new Promise(requestAnimationFrame);await new Promise(requestAnimationFrame)}')
panel=menu.locator('.composer-options-panel').bounding_box()
assert panel['x']>=0 and panel['x']+panel['width']<=page.locator('#composer-shell').bounding_box()['width']+1, panel
assert panel['y']>=0 and panel['y']+panel['height']<=900, panel
if menu_id != 'composer-options':
assert panel['y']+panel['height'] <= trigger.bounding_box()['y'], panel
page.screenshot(path=str(screenshots/f'{theme}-{width}-{menu_id}.png'),full_page=True,animations='disabled')
page.keyboard.press('Escape')
expect(menu).not_to_have_attribute('open', '')
expect(trigger).to_be_focused()
page.evaluate('()=>document.documentElement.dataset.theme="dark"')
page.locator('#composer-direction-trigger').click()
page.locator('#composer-reference-inputs-trigger').click()
expect(page.locator('#composer-direction')).not_to_have_attribute('open', '')
page.locator('#composer-instruction').click(position={'x':8,'y':8})
expect(page.locator('#composer-reference-inputs')).not_to_have_attribute('open', '')
# New conversations start with text, without modifying older conversations.
page.set_viewport_size({'width':1280,'height':900})
page.locator('#composer-new').click()
page.locator('#composer-attachment-files').set_input_files({'name':'test.png','mimeType':'image/png','buffer':__import__('base64').b64decode(PNG)})
expect(page.locator('#composer-send-images')).not_to_be_checked()
# A generation-only connection works without an understanding key.
from ai_provider_registry_e2e import open_ai_settings
library = open_ai_settings(run)
library.locator('[data-ai-routing-tab="providers"]').click()
library.locator('#open-ai-routing').click()
dialog = library.locator('#promptdirector-app-dialog')
dialog.locator('#promptdirector-app-dialog-providerEditor').select_option('custom-media')
expect(dialog.locator('#promptdirector-app-dialog-provider_custom_media_apiKey')).to_be_hidden()
key = dialog.locator('#promptdirector-app-dialog-provider_custom_media_imageApiKey')
key.click(); key.fill('generation-only-fixture')
dialog.locator('#promptdirector-app-dialog-provider_custom_media_imageProtocol').select_option('images_generations')
dialog.locator('#promptdirector-app-dialog-provider_custom_media_imageEndpoint').fill(origin+'/images/generations')
dialog.locator('#promptdirector-app-dialog-provider_custom_media_model_imageGeneration').fill('fixture-image-model')
dialog.locator('#promptdirector-app-dialog-provider_custom_media_consent').check()
library.screenshot(path=str(screenshots/'generation-connection.png'),full_page=True)
library.evaluate('()=>{chrome.permissions.request=async()=>true}')
dialog.get_by_role('button',name='保存配置',exact=True).click()
expect(dialog).to_be_hidden()
expect(library.locator('[data-provider-id="custom-media"]')).to_contain_text('已连接')
print(json.dumps({'referenceCombinations':4,'savedVisionModelSelectable':True,'narrowLayout':True, 'iconMenus':True, 'keyboardAndOutsideClose':True,'paidRequests':0}))
if __name__=='__main__': main()