Run the date parser first. Windows:

python "{baseDir}\scripts\parse_date_expr.py" --date-expr "<date_expr>"

If python is unavailable, retry with:

py -3 "{baseDir}\scripts\parse_date_expr.py" --date-expr "<date_expr>"

Run the resolver on the same expression.

python "{baseDir}\scripts\resolve_episodes.py" --date-expr "<date_expr>"

Use the resolver JSON as the source of truth for expanded_dates, episode_url, video_url, and target_filename.

Supported Date Expressions

Interpretation rules:

Default missing years to 2026
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:

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:

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.

Use the resolver's target_filename directly. Do not invent an alternate naming scheme.

Return one compact JSON-like summary in Markdown. Include these top-level fields:

summary must include:

Each item must include:

Status values for the final answer:

Map resolver failures into the final response like this:

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.