6.4 KiB
6.4 KiB
Fetch API Documentation
Fetch and cache API documentation in AI-readable markdown format.
Arguments
$ARGUMENTS
Library name, URL, or package identifier (e.g., "openai", "stripe", "https://docs.stripe.com")
Purpose
Download and cache API documentation locally for quick reference across all projects. Docs are stored globally in the plugin directory and available in all chat sessions.
Storage Location
Global (available across all projects):
~/.claude/plugins/marketplaces/unclecode-tools/plugins/unclecode-cc-toolkit/api-docs/
Execution Instructions
Step 1: Parse Input
Extract from query:
- Library name (e.g., "openai", "stripe", "nextjs")
- URL (if provided, e.g., "https://docs.stripe.com")
- Version (optional, e.g., "openai@4.0")
Normalize library name:
- Remove special characters
- Convert to lowercase
- Remove version suffix for directory name
Step 2: Check if Docs Exist
Check for existing documentation:
ls ~/.claude/plugins/marketplaces/unclecode-tools/plugins/unclecode-cc-toolkit/api-docs/{library}/
If directory exists:
- List what's there:
ls -lah api-docs/{library}/ - Show last modified date
- Ask user:
## Documentation Exists
**Library**: {library}
**Location**: api-docs/{library}/
**Last Updated**: {date from stat}
**Files**: {file count} files, {total size}
Documentation already cached. Would you like to:
1. Use existing docs (skip)
2. Update/re-fetch docs
3. View what's cached
Wait for user response.
If doesn't exist:
- Proceed to Step 3
Step 3: Fetch Documentation
Try these methods in order:
Method 1: Context7 API
If library name provided (no URL):
# Search for library
curl -X GET "https://context7.com/api/v1/search?query={library}" \
-H "Authorization: Bearer CONTEXT7_API_KEY"
# Get library ID from response
# Then fetch docs
curl -X GET "https://context7.com/api/v1/{library_id}/docs?type=txt&tokens=5000" \
-H "Authorization: Bearer CONTEXT7_API_KEY"
Note: Free tier available (no API key needed but rate-limited)
Save response to: api-docs/{library}/context7-docs.md
If successful, skip to Step 4.
Method 2: llms.txt
If URL provided OR Context7 doesn't have it:
- Extract domain from URL
- Try to fetch
{domain}/llms.txt:
curl -s https://{domain}/llms.txt
- If llms.txt exists:
- Parse the file (contains list of markdown doc URLs)
- Fetch each markdown URL listed
- Save each to:
api-docs/{library}/{filename}.md
If successful, skip to Step 4.
Method 3: Web Scraping
If neither Context7 nor llms.txt available:
-
Fetch the docs page HTML
-
Convert HTML to markdown using a clean method:
- Extract main content (remove nav, footer, ads)
- Convert to markdown
- Clean up formatting
-
Save to:
api-docs/{library}/scraped-docs.md
Add note at top of file:
<!-- Scraped from {URL} on {date} -->
<!-- May need manual review for accuracy -->
Step 4: Create Metadata File
Create api-docs/{library}/metadata.json:
{
"library": "library-name",
"source": "context7|llms.txt|web-scraped",
"sourceUrl": "original-url",
"fetchedAt": "2025-01-23T10:30:00Z",
"fileCount": 3,
"totalSize": "250KB",
"version": "4.0.0" (if specified)
}
Step 5: Create Index
If multiple files were fetched, create api-docs/{library}/INDEX.md:
# {Library} API Documentation
**Fetched from**: {source}
**Date**: {date}
**Version**: {version if known}
## Files
- [context7-docs.md](./context7-docs.md) - Main documentation
- [api-reference.md](./api-reference.md) - API reference
- [guides.md](./guides.md) - Usage guides
## Quick Links
{Parse and list major sections/topics}
Step 6: Confirm Success
Show summary:
## Documentation Fetched Successfully
**Library**: {library}
**Source**: {Context7 | llms.txt | Web}
**Location**: ~/.claude/plugins/.../api-docs/{library}/
**Files Saved**: {count} files ({total size})
### What's Available:
{List files with brief description}
### Usage:
These docs are now available globally across all projects.
To reference in your code:
- Ask Claude to "check {library} docs"
- Use /pp-triage for debugging with library context
- Docs automatically available in all future sessions
### Next Steps:
- View docs: cat api-docs/{library}/INDEX.md
- Update later: /pp-fetch-docs {library} (will prompt to update)
Query Examples
# Fetch by library name (tries Context7 first)
/pp-fetch-docs openai
/pp-fetch-docs stripe
/pp-fetch-docs nextjs
# Fetch by URL (tries llms.txt, then scrapes)
/pp-fetch-docs https://docs.stripe.com
/pp-fetch-docs https://platform.openai.com/docs
# Fetch specific version
/pp-fetch-docs openai@4.0
/pp-fetch-docs react@18
# Update existing docs
/pp-fetch-docs openai
> "Documentation exists. Update? (yes/no)"
> yes
> (re-fetches and overwrites)
Error Handling
If all methods fail:
## Unable to Fetch Documentation
**Library**: {library}
**Attempted**:
- ❌ Context7: Library not found
- ❌ llms.txt: Not available at {domain}
- ❌ Web scraping: {error reason}
### Suggestions:
1. Check library name spelling
2. Provide full URL to documentation
3. Try alternative library name (e.g., "@openai/sdk" vs "openai")
4. Manually download docs and save to:
~/.claude/plugins/.../api-docs/{library}/docs.md
Would you like me to:
- Search for alternative library names?
- Try a different URL?
- Help you manually organize docs?
Notes
- Docs are global - stored in plugin directory, not project-specific
- Available across ALL projects and chat sessions
- Context7 preferred (most up-to-date, version-specific)
- llms.txt is official standard (website-published)
- Web scraping is last resort (may need cleanup)
- Update anytime by re-running command
- Large docs may be split into multiple files for better organization
API Keys (Optional)
For Context7 higher rate limits:
- Get free API key: https://context7.com/dashboard
- Set environment variable:
export CONTEXT7_API_KEY="your-key" - Plugin will auto-detect and use it
Best Practices
When to fetch:
- Starting a new project with specific APIs
- Learning a new library
- Need up-to-date reference (re-fetch to update)
- Debugging issues (fresh docs help)
Storage management:
- Each library typically: 100-500KB
- Plugin directory has no size limit
- Delete old docs:
rm -rf api-docs/{library}/ - View all cached:
ls api-docs/