Knowledge

Add, review, and index the sources your organization's AI uses to answer questions

Knowledge is the organization-owned content its AI can retrieve when answering a visitor. Every source is saved as an editable draft first; review it, then index it before expecting the AI to use it.

Add a source

In the dashboard, open Harness → Knowledge and select Add source. Bubblo supports three source types.

Website

Enter either a page URL or a site root:

A full page URL, such as https://example.com/docs/refunds, is scraped immediately and saved as a draft.
A bare domain or root URL, such as example.com or https://example.com/, discovers up to 200 URLs through the site’s sitemap and internal links.

Site discovery creates URL entries only. It does not fetch every page. Open each discovered entry and select Scrape before indexing it.

File upload

Each upload can be up to 20 MB. The supported formats are:

Format	Extensions
PDF	`.pdf`
Microsoft Word	`.docx`
Markdown	`.md`, `.markdown`
Plain text	`.txt`

PDF parsing extracts embedded text. Bubblo does not perform OCR, so a scanned or image-only PDF with no extractable text cannot be used. Convert it to searchable text before uploading it.

Manual document

Enter a title and body directly. Manual documents are useful for short policies, FAQs, and content that does not already live in a file or on a public web page.

Review and index

Adding content and indexing it are separate steps:

Add a page, upload a file, or create a manual document.
For a URL found during site discovery, select Scrape to fetch that page.
Review the saved title and content. Edit them if needed.
Select Index.
Wait for each source to become Indexed.

Indexing runs asynchronously. The source remains in Indexing while Bubblo prepares and stores its searchable chunks; it is ready when the status changes to Indexed.

Source statuses

Status	Meaning	What to do
Not scraped	A URL was discovered, but its page content has not been fetched.	Open it and select Scrape.
Not indexed	The content is saved as a draft but the saved version is not yet in the current search index.	Review it, then select Index.
Indexing	The asynchronous indexing job is queued or running.	Wait for the status to update.
Indexed	The source is available to the organization’s knowledge retrieval.	Rebuild it only when you want to refresh the index.
Failed	The indexing job did not complete.	Review the error, correct the source if needed, and retry.

Edit or refresh a source

You can edit the title and content of a saved source. Website sources also have Re-scrape, which fetches that URL again and replaces the saved page content.

Editing or re-scraping returns the source to Not indexed. Select Index again and wait for indexing to finish. Until then, the latest saved changes are not available to knowledge retrieval.

Current limits

Only .pdf, .docx, .md, .markdown, and .txt uploads are supported.
The upload limit is 20 MB per file.
Image-only and scanned PDFs are not OCRed.
Site discovery returns at most 200 URLs, and every discovered page must be scraped individually.
Indexing is asynchronous; saving or scraping content alone does not index it.

Knowledge chunk and URL-scrape allowances vary by plan. See Billing and usage for the current limits. Once the content is indexed, test it in the widget and follow the resulting conversation in Inbox.