ascend-doc-auto-updater

name	ascend-doc-auto-updater
description	Detect and fetch the latest “Ascend C算子开发” PDF from the Huawei Ascend CANN Community Edition documentation, download page for a user-specified version, then convert the PDF into chapter-level markdown files and save them under ascend_dev_guide_sections/.
metadata	Auto-download AscendC operator dev PDF and export TOC-split Markdown.

Ascend Doc Auto Updater Skill

Purpose

When the user requests an update for a specific CANN Community Edition version (e.g. 850, 900beta1, 83RC1), this skill will:

1. Build the CANN documentation download page URL for that version

2. Run run.py locally with that URL

3. Let run.py automatically:

   • Locate the PDF URL for “Ascend C算子开发”
   • Download the PDF
   • Split content by TOC into section Markdown files
   • Write all Markdown files to ascend_dev_guide_sections/

Inputs (User Triggers)

The user may trigger this skill with phrases like:

• 更新到 850
• 更新版本：900beta1
• 拉取 83RC1 的 Ascend C算子开发文档
• 更新 CANN 社区版 850
• Or provide the full download page URL directly

Output Folder

• ascend_dev_guide_sections/ Each chapter/section is exported as one .md file.

Version Parsing Rules

1. If the user provides a full URL that matches:

   • .../CANNCommunityEdition/<version>/index/download then use it directly.

2. Otherwise extract <version> from the user message:

   • Prefer the token that looks like a CANN CE version identifier, such as:

     • 850, 860, 900beta1, 83RC1

   • If multiple different version tokens exist, choose the one that is:

     • Closest to keywords like CANN, CommunityEdition, 社区版, 版本, 更新
     • Otherwise choose the first best match

Download Page Construction

If a version token is extracted, construct:

download_page = https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/{version}/index/download

Execution Contract (Required Command)

When responding, the skill MUST provide:

1. The resolved version
2. The resolved download_page
3. A single shell command to run the update (include --force):

python run.py --download-page "<download_page>" --target-title "Ascend C算子开发" --force

Notes / Assumptions

• run.py must support CLI args:

  • --download-page (required)
  • --target-title (default "Ascend C算子开发")
  • --force (recommended to re-export even if the PDF URL/content is unchanged)

• The script is responsible for discovering the actual PDF URL from the download page and performing conversion.

• If the user input is ambiguous (e.g., multiple versions mentioned), prefer the version closest to “更新/版本/CANN/社区版” context; otherwise pick the first best match and proceed.

Examples

Example 1

User: 更新到 850 Skill output:

• version: 850
• download_page: https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/850/index/download
• command:

python run.py --download-page "https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/850/index/download" --target-title "Ascend C算子开发" --force

Example 2

User: 用这个链接更新：https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/900beta1/index/download Skill output:

• version: 900beta1
• download_page: https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/900beta1/index/download
• command:

python run.py --download-page "https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/900beta1/index/download" --target-title "Ascend C算子开发" --force