Fetch Webpage

Fetch webpages and convert them to clean Markdown format using a selectable Markdown proxy provider.

Usage

# Single URL, default provider is pure.md
node {baseDir}/scripts/fetch.mjs "https://example.com"

# Multiple URLs
node {baseDir}/scripts/fetch.mjs '["https://example.com", "https://example.org"]'

# With timeout
node {baseDir}/scripts/fetch.mjs "https://example.com" --timeout 60000

# Choose a specific provider
node {baseDir}/scripts/fetch.mjs "https://example.com" --provider defuddle.md

Options

• urls_json: JSON array of URLs or a single URL string
• --timeout <ms>: Request timeout in milliseconds (default: 30000)
• --provider <name>: One of pure.md, defuddle.md, markdown.new, r.jina.ai (default: pure.md)

Provider Selection

• If you already know a specific provider works better for this URL, domain, or content type, use that provider first
• Reuse the last successful provider for the same site within the current task when possible
• If there is no known preference, start with the default provider pure.md
• If the request fails or the Markdown quality is poor, retry the same URL with --provider defuddle.md
• If needed, retry again with --provider markdown.new
• Use --provider r.jina.ai as the last fallback
• Change only the --provider argument between retries unless you also need a longer --timeout

Guidance For AI

• Do not mechanically start from pure.md if the task context already indicates another provider is more reliable for that target
• Prefer explicit provider choice over blind retry loops
• When reporting what you did, mention which provider was used, especially if you skipped the default based on prior knowledge or prior failures

Notes:

• Returns webpage content converted to Markdown
• This script does not auto-fallback; the caller selects the provider explicitly
• For multiple URLs, returns a JSON array of results and includes the selected provider
• Node.js 18+ required (native fetch)