Skip to content

defuddle_fetch.md

Latest commit

 

History

History
38 lines (28 loc) · 1.35 KB

defuddle_fetch.md

File metadata and controls

38 lines (28 loc) · 1.35 KB
Raw

Defuddle Fetch (now web_fetch format=markdown)

Consolidated. The standalone defuddle_fetch tool was merged into web_fetch. Clean-markdown extraction is now a fetch mode, not a separate tool, so the model has one fewer tool to choose between.

Fetch cleaned markdown for a URL by calling web_fetch with format="markdown":

{
  "url": "https://example.com/complex-page",
  "format": "markdown"
}

This routes through the defuddle.md markdown extraction service and returns the cleaned markdown inline. Use it as a fallback when web_fetch (default format="summary") returns a payload that is hard to parse (heavy JS, paywalled HTML, raw RSS, etc.).

Optional parameters:

Field Type Default Description
url string — URL to fetch and extract
max_bytes number 256 KB Maximum response size

Guard Rails

  • Session cap: hard-capped at one call per session (the service is rate-limited)
  • Max bytes: 256 KB response limit
  • Timeout: 30s default, max 60s
  • Returns a structured error when the cap is hit, directing the agent back to web_fetch

Related Tools

  • web_fetch — fetch full page content (unlimited calls, more robust); format="markdown" gives defuddle extraction
  • web_search — search the web and get ranked result snippets