Skip to content

Webfetch Tool

Support web_fetch Tool

  • We want to support web_fetch tool for Keen. The idea is simple: it receives a URL and returns the content.
  • Can we consider converting pages to markdown? Is that feasible given that URLs may have javascript, html, images etc?
  • Save the plan in @.ai-interactions/tasks/issue-42/ as output-1_webfetch-tool.md
  • How many tokens in 512KB?
  • Is it not too big given that many LLMs have less than 300k context limit?
  • How many tokens in 128KB?
  • Which version of htmltomarkdown are you using? Why not the latest one?