The Knowledge page - every source your agent learns from

Written by Joseph Russo, Principal Architect, Networks and Tech, AskVault · Reviewed by Allison Chan

Last updated: June 18, 2026 · 5 min read

TL;DR

Knowledge is the single page where every source the bot learns from lives. URL crawls, file uploads, integrations, snippets, Q&A pairs. Monitor indexing status, re-sync sources and set each source's audience. Available on every plan, within your plan's content allowance.

What you see in Knowledge

A single sortable table of every knowledge source in your workspace. Each row carries:

• Source name. URL, filename, integration source, snippet title, or Q&A pair question.
• Type icon. URL, PDF, DOCX, Notion, GitHub, snippet, Q&A pair, etc.
• Status. Indexing, ready, failed, syncing, paused.
• Last sync timestamp.
• Audience. public or internal.
• Chunk count. How many vector chunks this source produces.
• Actions. Resync, edit, delete.

Above the table, a top bar shows aggregate stats: total sources, total chunks, storage used vs plan cap.

Source types

These source types appear in the same table:

• Web pages. Pages found from your site's sitemap, or pasted as exact links.
• File upload. PDF, TXT, MD and CSV on every plan; Word from Starter; more formats on higher plans (details).
• Imports. One-time imports from Notion, Confluence, GitHub, Zendesk, Salesforce, Slack and others.
• Snippet. Freeform text entries.
• Q&A pair. Curated question-answer entries.

There is no public API for adding knowledge today. All types share the same index, so retrieval pulls from any of them. The Audience setting works uniformly across types.

Add a new source

The + Add Source button in the top right opens the source-type picker. Pick the type you want, follow the per-type setup flow.

Most common starting flows:

• Web Crawler. Paste a URL (Domain Crawl) or a list of pages (Exact Links), click Start.
• File upload. Drag-and-drop one or more files.
• Notion import. Connect Notion, then import the pages you shared.
• Q&A pair. Type the question and answer directly.

Detailed setup steps for each type live in the knowledge management cluster.

Monitor indexing status

When you add a source, it goes through:

1. Queued. Waiting to start.
2. Indexing. Crawling, extracting, chunking, embedding.
3. Ready. Available for retrieval. Bot can cite from it.
4. Failed. Something went wrong. Click for error details.

Status updates in real time. When a resync finds a page changed, the page is queued and indexed again; your agent answers from the new text once it's ready, and the page may be briefly unavailable while it re-indexes.

For sources with retry budgets (transient failures), AskVault auto-retries on an exponential schedule. Permanent failures (such as a 404) don't retry; click to see the specific failure code.

Trigger a re-sync

To pull the latest content from a source:

• One page: open the row's Source Options menu and click Force Resync.
• Several pages: select them and click Resync selected in the selection bar, or Resync all for every source with a web address in the workspace.

Resync checks each page for changes and re-reads only the ones that changed. Paid plans can set an automatic re-crawl schedule, but until automatic re-crawl is on for your account a scheduled check may not re-read a changed page, so keep using Resync to be sure (details). See keeping website knowledge up to date for what resync does and doesn't cover.

Resync all also re-reads your imports

Resync all includes sources imported from Notion, Zendesk, Confluence, Salesforce, Slack and similar, and re-reads them as if they were public web pages. That can mark them as needing attention, or replace the imported text with whatever the public page shows. If you have imports, use Resync selected on your website pages, and refresh imports by running the import again.

Bulk actions

Multi-select sources (Shift+click or checkbox in the leftmost column), then apply:

• Delete. Removes the selected sources and their content from your agent's knowledge.
• Resync selected. Checks the selected web pages for changes and re-reads the changed ones.
• Resync all. The same, for every source with a web address in the workspace, including imported sources (see the caution above).

Storage usage indicator

The top bar shows current storage usage vs your plan cap:

• Used. Total MB across all sources.
• Cap. 5 MB on Free, 15 MB on Starter, 40 MB on Growth, 100 MB on Business.
• Percentage. Visual bar.

At 100% of your content cap, new content is not indexed until you delete something or upgrade. Existing content keeps working. Check usage in the usage meter.

Audience at the source level

Click any source to expand its detail and set its Audience. It has exactly two values:

• public: the default. Eligible for every channel, including the widget and the hosted page.
• internal: excluded from retrieval whenever the request runs on a public channel.

There are no other audience values, and there is no bulk audience control. See Audience on knowledge sources.

Chunk inspection

For debugging "why is the bot answering wrong?", inspect individual chunks:

1. Click a source row to expand.
2. View Chunks opens a paginated list of every chunk extracted from this source.
3. Each chunk shows token count, parent heading prefix, and the raw text.

Useful for spotting chunking issues, missing context, or content that didn't extract correctly.

Search across sources

The Knowledge search bar searches both source names and chunk content:

• By name. "shopify" finds the Shopify integration source.
• By content. "refund policy" finds any chunk containing those words.

Search results show the source plus highlighted matching chunks.

Limits

• Plan availability. Knowledge on every plan including Free.
• Source count. No hard cap; your plan's content allowance is the limit.

Common pitfalls

Bot answers wrong despite the right content existing. Check Knowledge > [source] > Chunks to confirm the right text is indexed. If chunking is off, re-upload as a different format (e.g., DOCX instead of PDF) or adjust the source.

Sync stuck on "Indexing" for hours. Source has parsing issues. Click for error details. Common causes: corrupted PDF, image-only PDF without OCR, password-protected file.

Internal content answered on the widget. The source's Audience is still public (the default). Set it to internal on the source that owns the content: it is not inherited from the source it was crawled or imported from.

Storage cap hit unexpectedly. Deleted sources can take a short while to free their space. Refresh after a few minutes.

FAQ

Can I export everything in the Knowledge page?

No. There is no bulk export of your knowledge today. Keep the original files and pages you imported.

How do I know which source the bot cited in a conversation?

Open the conversation in Inbox > Chat History. Each bot message lists sources used. Click a source to jump to its row in Knowledge.

Can I share sources between workspaces?

Not directly. Each workspace has its own Knowledge. For shared knowledge, use an integration that both workspaces ingest from (e.g., the same Notion workspace).

Does the bot prefer some sources over others?

Q&A pairs get a retrieval boost, so a close match usually ranks first. There is no setting to weight other sources.

What happens to deleted sources?

Deleted content is removed from search right away and cannot be restored from the dashboard.

Related guides

• URL crawling
• PDF and document uploads
• Snippets
• Q&A pairs
• Keeping website knowledge up to date

See RAG-grounded answers in action

AskVault answers from the content you index, shows its sources by default, and says so when it finds nothing.

Try AskVault Free → Read the API docs

Was this page helpful? Yes No