Importing knowledge: websites, PDFs, folders and GitHub
You rarely start from nothing. Starko imports the documentation you already have, keeps the structure where it can, and embeds everything for the support assistant. Press Create article in the Knowledge Hub and the dialog offers three ways in: Create new article, Import from website and Import files.
kh create article
From a website
Paste the URL of your help center or documentation site. Starko crawls the pages under it, creates one article per page with headings, lists, links and images preserved, and files them in the collection and subcollection you choose. Re-running the import on the same site updates existing articles rather than duplicating them.
Good for: an existing help center, product docs, FAQ pages, pricing and policy pages.
From files
Choose Import files and upload one or many PDFs. Each becomes an article with the text extracted and chunked for search. Scanned documents are read with OCR. Choose a collection and subcollection for the batch.
Good for: manuals, contracts and policies, slide decks exported as PDF.
From a folder
In the same Import files dialog, drop a whole folder of Markdown, text, Word or HTML files. The folder structure becomes collections and subcollections, and each file becomes an article. Duplicate titles are merged.
Good for: a docs folder from a repository, an export from Notion or Confluence.
From GitHub
Under Knowledge Hub → Skills → Import from GitHub, enter owner/repo to import skills in the skills//SKILL.md layout. They arrive as drafts; publish the ones you want.
From a conversation
In Ask Starko (⌘L) or Slack, ask Starko to write an article from a web page, a ticket or a conversation. It drafts the article as a private draft for you to review and publish.
After importing
Open the collection and skim the titles. Rename anything that reads like a file name.
Set visibility. Imports arrive private so nothing goes live by accident.
Check Topic Gaps a week later to see what the imported content still does not cover.
Limits
Articles are capped at 30,000 characters each; very long pages are split. Binary content in PDFs (images without text) is not searchable.
Did this answer your question?