/url-or-path

Import URL or Path

You are executing the /import:url-or-path command. Import documentation from a URL or local path.

Arguments

• $ARGUMENTS - URL (https://...) or local file/directory path

Dynamic Folder Structure

All imports go to mabi-import/ with intelligent substructure:

Documentation Sites (detected by Context7 or docs.* domain)

mabi-import/{library}/{version}/
  datahub.md
  data-importer.md

Example: https://docs.pimcore.com/platform/Data_Importer/ (version 2025.4) -> mabi-import/pimcore/v2025.4/data-importer.md

Regular Websites

mabi-import/website/{domain}/
  {page-name}.md

Example: https://example.com/api/docs -> mabi-import/website/example.com/api-docs.md

Local Files

mabi-import/local/{project-or-context}/
  {filename}

Determine context from:

• Parent directory name (if meaningful, not generic like "Downloads")
• File content (check for project name in headers)
• Ask user if unclear

Example: /home/user/projects/myapp/docs/config.md -> mabi-import/local/myapp/config.md

Process

1. Determine Source Type

Check if $ARGUMENTS is:

• URL: Starts with http:// or https://
• Local path: Everything else (file or directory)

2a. Local Path Import

If local path:

1. Check if exists:

   ls -la "$ARGUMENTS"
   

2. Determine context for folder structure:

   • Extract project name from path (e.g., /projects/myapp/ -> myapp)
   • If path is generic (Downloads, tmp, home), ask user for context name
   • Read file header if available to detect project/tool name

3. Confirm with user:

   You specified a local path: {path}
   
   Suggested location: mabi-import/local/{context}/{filename}
   
   Is this correct? (The path will be copied, not moved)
   

4. Use AskUserQuestion to confirm or let user provide different context.

5. If confirmed:

   mkdir -p "mabi-import/local/{context}"
   cp -r "$ARGUMENTS" "mabi-import/local/{context}/"
   

2b. URL Import - Priority Chain

If URL, try methods in this order. Inform user at each step:

Step 1: Context7 (fastest, structured)

1. Extract library name from URL (e.g., "pimcore" from docs.pimcore.com)
2. Use mcp__plugin_context7_context7__resolve-library-id to check if library exists
3. If found:
   • Inform user: "Found in Context7! Fetching structured documentation..."
   • Use mcp__plugin_context7_context7__query-docs to get relevant content
   • Detect version from Context7 response or URL
   • Save to mabi-import/{library}/{version}/
   • Done!
4. If NOT found or query fails:
   • Inform user: "Not found in Context7. Trying WebFetch..."
   • Continue to Step 2

Step 2: WebFetch (fast, direct)

1. Use WebFetch tool with the URL
2. If successful (status 200, content returned):
   • Detect if documentation site (domain contains "docs", "developer", "api")
   • If docs: Try to extract library name and version -> mabi-import/{library}/{version}/
   • If website: Use domain -> mabi-import/website/{domain}/
   • Done!
3. If fails (403, blocked, captcha, timeout):
   • Inform user: "WebFetch blocked (403/captcha). Trying Playwright..."
   • Continue to Step 3

Step 3: Playwright (slow, but bypasses blocks)

1. Inform user: "Using Playwright browser..."
2. Use Playwright MCP tools:
   • browser_navigate to the URL
   • browser_snapshot to get page content
   • Extract version from page if visible (look for version selectors, breadcrumbs)
   • Extract relevant documentation
3. Determine folder structure same as WebFetch
4. Close browser when done
5. If Playwright also fails:
   • Inform user: "All methods failed. Cannot fetch this URL."
   • Suggest: Check if URL is correct, try manual copy, check network

IMPORTANT: Do NOT use headless mode. Many sites (especially those using Cloudflare) detect and block headless browsers. The visible browser window is intentional - it allows passing bot detection and manual captcha solving if needed.

3. Version Detection (REQUIRED)

Try to detect version from:

1. URL path: /v2/, /v2025.4/, /version/19/
2. Page content: Version selector, breadcrumb, title
3. Context7 metadata: Version info in response (check for version field or version in library ID)

IMPORTANT: Never use "latest" as version folder name. If no version can be detected:

• Ask user: "No version detected. What version is this documentation for?"
• User must provide version (e.g., "v2025.4", "v19", "v3.2.1")
• Only proceed after version is confirmed

Exception: For website/ and local/ categories, version is not required (content is not versioned).

4. Save Documentation

Target location based on type:

• Docs: mabi-import/{library}/{version}/{page-name}.md
• Website: mabi-import/website/{domain}/{page-name}.md
• Local: mabi-import/local/{context}/{filename}

Format saved file:

---
source: {original URL or path}
imported: {timestamp}
method: {context7|webfetch|playwright|local}
version: {exact version - NEVER "latest"}  # omit for local/website
---

# {Title}

{Content}

Note: version field is only required for library documentation. Omit it for local/ and website/ imports.

5. Output

On success:

Documentation imported:

  Source:  {url or path}
  Method:  {context7|webfetch|playwright|local}
  Version: {version}
  Saved:   mabi-import/{structure}/{name}.md
  Size:    {file size}

Use /import:list to see all cached docs.
Use /import:search <query> to search within docs.

On complete failure:

Failed to import documentation:

  Source: {url or path}

  Tried:
    1. Context7:   {not found / error}
    2. WebFetch:   {403 blocked / error}
    3. Playwright: {error}

Cannot fetch this URL automatically.

Suggestions:
- Check if URL is correct and accessible
- Copy content manually and use /import:url-or-path /local/path
- Check if Context7/Playwright plugins are installed

Notes

• Priority: Context7 -> WebFetch -> Playwright
• User is informed at each fallback step
• Playwright runs visible (NOT headless) to bypass bot detection
• Local paths are copied, not moved
• Always ask for confirmation on local paths
• Saved docs include metadata (source, method, version, import date)
• Folder structure keeps imports organized by source type and version