Find 安徽网络电视台《快乐无敌大PK》 full-episode pages from a user-provided date expression, extract each episode's real video URL, and save the episodes into 迅雷云盘. Use th...
For 3月5、6、7、8日, inherit month and year from the first date
For 3月5-10日, inherit month and year for the end date from the start date
For 3月5日之后 and 3月5日以后, exclude the day itself and start from 2026-03-06
For open ranges, stop at the latest currently discoverable 快乐无敌大PK episode date on 安徽网络电视台
Reject these requests as unsupported instead of guessing:
起, 以来, 前后, 本周, 最近几天
Cross-year shorthand without an explicit year
Requests that mix 快乐无敌大PK with a different program
Resolution Workflow
Run the resolver before opening 迅雷云盘. The resolver already does the deterministic parts:
Search https://www.ahtv.cn/search for 快乐无敌大PK
Filter candidates to https://www.ahtv.cn/pindao/ahzh/pk/split/...
Match by the episode date embedded in the title, not by the search result publish time
Prefer titles containing 整期
Cross-check ambiguous dates against https://www.ahtv.cn/pindao/ahzh/pk and its paginated index pages
Extract the real media URL from the single-episode page hidden input #m3u8
Treat the resolver output like this:
status=ready: proceed to 迅雷云盘
status=not-found: record the failure and continue with the next date
status=ambiguous: record the failure and continue with the next date
status=video-url-missing: record the failure and continue with the next date
status=error: record the failure and continue with the next date
迅雷云盘 Workflow
Process dates one by one and continue even if one date fails.
For every resolver item with status=ready:
Open https://pan.xunlei.com/.
If the user is not logged in, use $sms-login-dom-first to complete 迅雷云盘 login with phone number plus SMS code, then return to the cloud drive page.
Search the whole cloud drive for the exact target_filename.
If an exact same file already exists anywhere in 迅雷云盘, mark the item as skipped-existing and do not add it again.
If it does not exist, ensure the folder path /快乐无敌大PK/2026 exists.
Use 迅雷云盘's link-based add flow. UI labels can vary, so look for controls such as 链接添加, 添加链接, 云添加, 添加任务, or a plus menu with a chain/link icon.
Paste the resolver video_url exactly as returned.
Save into /快乐无敌大PK/2026 if the dialog lets you choose the destination. If it saves elsewhere first, move it into /快乐无敌大PK/2026 immediately after creation.
Wait until the file becomes visible in the target folder.
If 迅雷 keeps the source file name, rename it to target_filename.
Do not download to local disk as a fallback in v1. Only use 迅雷云盘 link add / cloud add.
Naming Rules
Basename format: 快乐无敌大PK.2026.S02E{MMDD}
Final filename format: 快乐无敌大PK.2026.S02E{MMDD}.mp4
Example: 快乐无敌大PK.2026.S02E0315.mp4
Use the resolver's target_filename directly. Do not invent an alternate naming scheme.
Final Output Format
Return one compact JSON-like summary in Markdown. Include these top-level fields:
input_expr
expanded_dates
summary
items
summary must include:
total
added
skipped_existing
not_found
failed
Each item must include:
date
status
episode_url
video_url
target_filename
xunlei_path
message
Status values for the final answer:
added-and-renamed
skipped-existing
not-found
add-failed
Map resolver failures into the final response like this:
not-found stays not-found
ambiguous, video-url-missing, and error become add-failed
Notes
Prefer the resolver scripts over ad-hoc parsing in the model.
Treat the site structure as authoritative as of the current run. If the resolver fails because the site changed, report the failure clearly instead of fabricating URLs.
Keep the browser workflow deterministic: exact filename search first, then add, then rename if needed.
