Skip to content

Knowledge base

The bot answers from three places:

  • the knowledge base: your website, documents, text and FAQ, searched for every reply;
  • the business profile: a Markdown summary of your business that the bot always knows;
  • custom instructions: tone of voice, things to avoid, how to handle edge cases.

Add sources in Bot → Knowledge.

Source Details
Website A page URL. Choose how much of the site to import (see Website scope). Pages that need JavaScript to show their content (single-page apps) are rendered in a browser, and robots.txt is respected.
Sitemap A sitemap.xml URL: Feedbot imports the pages listed in it. Choose Pages listed in a sitemap.xml when adding a website (it is picked for you when the URL looks like a sitemap).
Files PDF, DOCX, Markdown, TXT or CSV, up to 15 MB each. Text longer than 2 MB is shortened.
Text Any text, with a title, up to 200,000 characters.
FAQ Questions and answers in the form Q: ... / A: ....

When you add a website, choose what to import:

Option What Feedbot imports
Only this page (default) Just the URL you entered. Good for a single FAQ, pricing or docs page.
This page and the pages it links to The page plus the pages on the same site it links to, up to the depth you pick: 1 level is the page and the pages linked from it, 2 levels adds the pages linked from those, 3 levels goes one step further.
Whole website The site’s sitemaps and every page linked on the same domain.

Every option is limited only by your plan’s remaining knowledge pages; a single website can use all of them. Each website source in the list shows its scope, for example Only this page, Linked pages · depth 2 or Whole website. To change the scope, delete the source and add the website again.

Websites are not re-crawled automatically. After you change your site, click Refresh website: Feedbot re-reads only the pages that changed since the last crawl, within the source’s scope, and keeps the rest. The source shows when it was last refreshed. A source shows Truncated when it was cut off by a limit.

Every source is split into passages of about 2,400 characters, each prefixed with its nearest heading. For each reply, Feedbot runs a semantic (embedding) search and a keyword search, merges the results, and gives the bot the 5 most relevant passages. The business profile and custom instructions are included in every reply.

So keep each page focused, use clear headings, and put facts the bot needs in every conversation (prices, contacts, opening hours) in the business profile.

Knowledge limits apply to the whole workspace, across all bots.

Plan Websites Knowledge pages Characters Files
Free — — (text & FAQ only) 20,000 —
Starter 1 100 1,000,000 ✓
Pro 3 500 5,000,000 ✓
Business 10 2,000 20,000,000 ✓

On Free, the knowledge base is Text and FAQ only. Website import and file uploads start on Starter.

When the character limit is reached, website imports are cut off and new text, FAQ or files are rejected. Delete sources you don’t need or upgrade.

Edit both in Bot → Behavior. Maximum length in characters:

Plan Custom instructions Business profile
Free 3,000 5,000
Starter 6,000 10,000
Pro 10,000 20,000
Business 20,000 40,000

Feedbot’s own platform rules and role modules (support, sales, feedback) are always included. Your instructions are added on top.

The bot always answers in the language the visitor writes in. Main language in Bot → Behavior is used when that is unclear and for the widget interface. With Auto, it is the visitor’s browser language.